SOLUTION MANUAL
PROBLEM 1.1
FIND: Explain the hierarchy of standards. Explain the term standard. Cite example.
SOLUTION
The term standard refers to an object or instrument, a method or a procedure that provides a
value of an acceptable accuracy for comparison.
A primary standard defines the value of the unit to which it is associated. Secondary
standards, while based on the primary standard, are more readily accessible and amenable
for use in a calibration. There is a hierarchy of secondary standards: A transfer standard
might be maintained by a national standards lab (such as NIST in the United States) to
calibrate industrial “laboratory standards”. It is costly and time-consuming to certify a
laboratory standard, so they are treated carefully and not used too regularly. A laboratory
standard would be maintained by a company to be used to certify a more common in-house
reference called the working standard. A working standard would be calibrated against the
laboratory standard. The working standard is used on a more regular basis to calibrate
everyday measurement devices or products being manufactured. Working standards are more
the norm for most of us. A working standard is simply the value or instrument that we
assume is correct in checking the output operation of another instrument.
Example: A government lab maintains the primary standard for pressure. It calibrates a an
instrument called a “deadweight tester” (see C9 discussion) for high pressure calibrations.
These form its transfer standard for high pressure. A company that makes pressure
transducers needs an in-house standard to certify their products. They purchase two
deadweight testers. They send one tester to the national lab to be calibrated; this becomes
their laboratory standard. On return, they use it to calibrate the other; this becomes their
working standard. They test their manufactured transducers using the working standard –
usually at one or two points over the transducer range to assure that it is working. Because
the working standard is being used regularly, it can go out of calibration. Periodically, they
check the working standard calibration against the laboratory standard.
See ASME PTC 19.2 Pressure Measurements for a further discussion.
A test standard defines a specific procedure that is to be followed.
PROBLEM 1.2
FIND: Why calibrate? What does calibrated mean?
SOLUTION:
The purpose of a calibration is to evaluate and document the accuracy of a measuring device.
A calibration should be performed whenever the accuracy level of a measured value must be
ascertained.
An instrument that has been calibrated provides the engineer a basis for interpreting the
device’s output indication. It provides assurance in the measurement. Besides this purpose, a
calibration assures the engineer that the device is working as expected.
A periodic calibration of measuring instruments serves as a performance check on those
instruments and provides a level of confidence in their indicated values. A good rule is to
calibrate measuring systems annually or more often as needed.
ISO 9000 certifications have strict rules on calibration results and the frequency of
calibration.
PROBLEM 1.3
FIND: Suggest methods to estimate the accuracy and random and systematic errors of a
dial thermometer.
SOLUTION:
Random error is related to repeatability: how closely an instrument indicates the same value.
So a method that repeatedly exposes the instrument to one or more known temperatures
could be developed to estimate the random error. This is usually stated as a statistical
estimate of the variation of the readings. An important aspect of such a test is to include
some mechanism to allow the instrument to change its indicated value following each
reading so that it must readjust itself.
For example, we could place the instrument in an environment of constant temperature and
note its indicated value and then move the instrument to another constant temperature
environment and note its value there. The two chosen temperatures could be representative of
the range of intended use of the instrument. By alternating between the two constant
temperature environments, differences in indicated values within each environment would be
indicative of the precision error to be expected of the instrument at that temperature. Of
course, this assumes that the constant temperatures do indeed remain constant throughout the
test and the instrument is used in an identical manner for each measurement.
Systematic error is a fixed offset. In the absence of random error, this would be how closely
the instrument indicates the correct value. This offset would be present in every reading. So
an important aspect of this check is to calibrate it against a value that is at least as accurate as
you need. This is not trivial.
For example, you could use the ice point (0oC) as a check for systematic error. The ice point
is formed from a carefully prepared bath of solid ice and liquid water. As another check, the
melting point of a pure substance, such as silver, could be used. Or easier, the steam point.
Accuracy requires a calibration to assess both random and systematic errors. If in the
preceding test the temperatures of the two constant temperature environments were known,
the above procedure could serve to establish the systematic error, as well as random error of
the instrument. To do this: The difference between the average of the readings obtained at
some known temperature and the known temperature would provide an estimate of the
systematic error.
PROBLEM 1.4
FIND: Discuss interference in the test of Figure 1.3
SOLUTION :
In the example described by Figure 1.3, tests were run on different days on which the local
barometric pressure had changed. Between any two days of different barometric pressure, the
boiling point measured would be different – this offset is due to the interference effect of the
pressure.
Consider a test run over several days coincident with the motion of a major weather front
through the area. Clearly, this would impose a trend on the dataset. For example, the
measured boiling point may be seem as increasing from day to day.
By running over random days separated by a sufficient period of days, so as not to allow any
one atmospheric front to impose a trend on the data, the effects of atmospheric pressure can
be broken up into noise. The measured boiling point might then be high one test but then low
on the next, in effect, making it look like random data scatter, i.e. noise.
PROBLEM 1.5
FIND: How does resolution affect accuracy?
SOLUTION
The resolution of a scale is defined by the least significant increment or division on the
output display. Resolution affects a user's ability to resolve the output display of an
instrument or measuring system.
Consider a simple experiment to show the effects of resolution. Under some fixed condition,
ask several competent, independent observers to record the indicated value from a
measurement system. Collect the results – this becomes your dataset. Because the indicated
value is the same for each observer, the scatter in your dataset will be close to the value of
the resolution of the measurement system.
Data scatter contributes to the random error. As such, the output resolution of a measurement
system forms a lower limit as to the random error to be expected.
The resolution would not contribute to systematic error. Systematic error is an offset.
PROBLEM 1.6
FIND: How does hysteresis affect accuracy?
SOLUTION
Hysteresis error is the difference between the values indicated by a measurement system
when the value measured is increasing in value as opposed to when it is decreasing in value;
this despite the fact that the value being measured is actually the same for either case.
A common cause of hysteresis in analog instruments is friction in the moving parts. This can
cause the output indicator to 'stick'. In digital instruments, hysteresis can be caused by the
discretization.
If hysteresis error is ignored, the effect on any single measurement can be seen as a
systematic error. On multiple measurements in any one direction, the effect can be to impose
a 'trend' on the data set. The use of randomization methods can break up the trends
incorrectly implied by hysteresis effects. Randomization makes systematic errors behave as
random errors, which are more easily interpreted. If randomization methods are not used, the
hysteresis effect behaves as a systematic error.
PROBLEM 1.7
SOLUTION
This problem is open-ended and has no unique solution. We suggest that the instructor use
this Problem as the basis for an in-class or small group discussion.
PROBLEM 1.8
FIND: Identify measurement stages for each device.
SOLUTION
a)
thermostat
Sensor/transducer: bimetallic thermometer
Output: displacement of thermometer tip
Controller: mercury contact switch (open:furnace off; closed:furnace on)
b)
speedometer
Method 1:
Sensor: usually a mechanically coupled cable
Transducer: typically a dc generator that is turned by the cable producing an electrical
signal
Output: typically a pointer/scale (note: often a galvanometer is used to convert the
electrical signal in a mechanical rotation of the pointer)
Method 2:
Sensor: A magnet attached to the rotating shaft
Transducer: A Hall Effect device that is stationary but detects each sensor passage by
creating voltage pulse
Signal Conditioning: A pulse counting circuit; maybe also digital-analog converter (if
analog readout is used)
Output: An analog or digital readout calibrated to convert pulses per minute to kph or
mph.
c)
Portable CD Stereo Player
Sensor: laser with optical reader (reflected light signal differentiates between a "1" and
"0")
Transducer: digital register (stores digital information for signal conditioning)
Signal conditioning: digital-to-analog converter and amplifier (converts digital numbers
to voltages and amplifies the voltage)
Output: headset/speaker (note: the headeset/speaker is a second transducer in this system
converting an electrical signal back to a mechanical displacement)
d)
anti-lock braking system
Sensor: brake activation switch senses brakes 'on'; encoder counts wheels revolutions per
unit time
Signal conditioning: timing circuit
Output: a feedback signal that pulsates brake action overriding the driver's constant pedal
pressure
e)
audio speaker
Sensor: coil (to which the input terminals are connected)
Transducer: coil-magnet-speaker cone that acts as a miniature electrical dc motor
responding to changes in current applied.
Output: speaker cone displacement
PROBLEM 1.9
KNOWN: Data of Table 1.5
FIND: input range, output range
SOLUTION
By inspection
0.5 ≤ x ≤100 cm
0.4 ≤ y ≤ 253.2 V
The input range (x) is from 0.5 to 100 cm. The output range (y) is from 0.4 to 253.2V. The
corresponding spans are given by
ri = 99.5 cm
ro = 252.8 V
COMMENT
Note that each answer has units shown. By themselves, numerical answers are meaningless.
Always show units for data, for each step of data reduction and in all reported results.
PROBLEM 1.10
KNOWN: Data set of Table 1.5
FIND: Discuss advantages of different plot formats for this data
SOLUTION:
Both rectangular and log-log plots are shown below.
Rectangular grid (left plot below):
An advantage of this format is that is displays the data clearly as having a non-linear
relationship. The data trend, while not immediately quantifiable, is established.
A disadvantage with this data set is that the poor resolution at low x values makes
quantification at low values difficult.
Log-log grid (right plot below):
An advantage of this format with this particular data set is that the data display a linear
relationship of the form: log y = m log x + log b. This tells us that the data have the
relationship, y = bxm. Because of these facts, resolution is equally good over the whole scale.
A disadvantage with this format is that one must remember the data has been conditioned to
look linear. We are no longer plotting x versus y. This is particularly important to remember
when attempting to find the slope of y against x.
P1.10
P1.10
300
1000
250
100
Y[V]
Y[V]
200
150
100
10
1
50
0.1
0
0
50
X[cm]
100
0.1
1
10
X[cm]
100
PROBLEM 1.11
KNOWN: Calibration data of Table 1.5
FIND: K at x = 5, 10, 20 cm
SOLUTION:
The data reveal a linear relation on a log-log plot suggesting y = bxm. That is:
log y = log (bxm) = log b + m log x
Y = B + mX
or
From the plot, B = 0, so that b = 1, and m = 1.2. Thus, we find from the calibration the
relationship
y = x1.2
Because K = [dy/dx]x = 1.2x0.2, we obtain
P1.11
K [V/cm]
1.66
1.90
2.18
1000
100
Y[V]
x [cm]
5
10
20
We should expect that errors would
propagate with the same sensitivity as the
data. Hence for y=f(x), as sensitivity
increases, the influence of the errors on y
due to errors in x between would increase.
10
1
0.1
0.1
1
10
100
X[cm]
COMMENT
A common shortcut is to use the approximation that
dy/dx = lim x→0 ∆y/∆x
This approximation is valid only for very small changes in x, otherwise errors result. This is a
common mistake. An important aspect of this problem is to draw attention to the fact that
many measurement systems may have a static sensitivity that is dependent on input value.
PROBLEM 1.12
KNOWN: Sequence calibration data set of Table 1.6
ri = 5 mV
ro = 5 mV
FIND:
%(eh)max
SOLUTION
By inspection of the data, the maximum hysteresis occurs at x = 3.0. For this case,
eh = (eh)max = yup - ydown
= 0.2 mV
or
%(eh)max = 100 x (0.2 mV/5 mV)
= 4%
P1.12
5
Y[mV]
4
3
2
1
0
0
2
4
X[mV]
Problem 1.13
KNOWN: Comparison of three clock outputs with standard time
FIND: Discuss estimated accuracy
SOLUTION
Clock A shows a bias error of 2:23 s. The bias would appear to be increasing at a rate of 1
s/hr. However, clock resolution is 1 s which by itself can lead to precision error (data scatter)
of ± ½ s; this can create the situation noted here. Another reading would clarify this.
Clock B shows a bias error of 5 s. There does not appear to be any precision error in the
output.
Clock C shows a 0 s bias error and a precision error on the order of ± 2 s.
Because of the calibration, we now know the values of bias error for each clock. Correcting
for bias error, we can consider Clock B to provide the more accurate time. Over time, the
bias error in Clock A could become cumbersome to deal with, that is if the bias is indeed
increasing in time. Therefore, it provides the least reliable value of time.
PROBLEM 1.14
SOLUTION
Each curve is plotted below in a suitable format to yield a linear shape.
(a)
100
(b)
14
12
1
10
0.1
8
Y
Y[V]
10
6
0.01
4
0.001
0.0001
0.01
2
0
0.1 X[m]
1
10
0
2
4
6
8
10
(d)
(c)
10
X
1.E+05
1.E+03
Y
1.E+01
Y
1
1.E-01
1.E-03
1.E-05
1.E-07
0.1
1
X
10
0.01
0.1
(e)
1.E+02
1.E+00
1.E-02
Y
0.1
0.01
1.E-04
1.E-06
1.E-08
0
2
4
X
6
8
10
X
1
10
PROBLEM 1.15
KNOWN: y = 10e-5x
FIND: Slope at x = 0, 2 and 20
SOLUTION
The equation has been plotted below. The slope of the equation at any value of x can be
found graphically or by the derivative
dy/dx = -50e-5x
x [V] dy/dx [V/unit]
0
2
20
-50
-0.00227
0
The sensitivity of y to x decreases with x.
COMMENT
An important aspect of this problem, is to draw attention to the fact that many measurement
systems may have a static sensitivity that is dependent on input value. While it is desirable to
have a constant K value, the operating principle of many systems will preclude this or
incorporate signal conditioning stages to overcome such nonlinearity. In Chapter 3, the
concept that system's also have a dynamic sensitivity that is frequency dependent will be
introduced.
P1.15
y [V]
10
1
0.1
0.01
0
10
X [units]
20
PROBLEM 1.16
SOLUTION
The data are plotted below. The slope of a line passing through the data is 1.365 and the y
intercept is 2.12. The data can be fit to the line y = 1.365x +2.12. Therefore, the static
sensitivity is K = 1.365 for all x.
P1.16
10
Y [V]
8
6
4
2
0
0
2
X [V]
4
6
PROBLEM 1.17
KNOWN: Data of form y = axb.
FIND: a and b; K
SOLUTION
The data are plotted below. If y = axb, then in log-log format the data will take the linear
form
log y = log a + b log x
A more or less linear curve results with this data. From the plot, the curve fit found is
log y = -0.23 + 2x
This implies that
y = 0.59x2
so that a = 0.59 and b = 2. The static sensitivity is found by the slope dy/dx at each value of
x.
x [m]
0.5
2.0
5.0
10.0
K(x1) = dy/dx x1 [cm/m]
0.54
2.16
5.40
10.80
COMMENT
An aspect of this problem is to draw attention to the fact that many measurement systems
may have a static sensitivity that is dependent on input value. The operating principle of
many systems will determine how K behaves.
p1.17
y [cm]
100
10
1
0.1
0.1
1
X [m]
10
PROBLEM 1.18
KNOWN: Calibration data
FIND: Plot data. Estimate K.
SOLUTION
The data are plotted below in semi-log format. A linear curve results.
This suggests y = aebx. Plotting y vs x in semi-log format is equivalent to
plotting
log y = log a + bx
From the plot, a = 5 and b = -1. Hence, the data describe y = 5e-x. Now, K =
dy/dx x, so that
X [psi]
K
0.05
0.1
0.5
-4.76
-4.52
-3.03
1.0
-1.84
The magnitude of the static sensitivity decreases with x. The negative sign indicates that y
will decrease as x increases.
COMMENT
An aspect of this problem is to draw attention to the fact that many measurement systems
may have a static sensitivity that is dependent on input value. The operating principle of
many systems will determine how K behaves.
P1.18
10
Y [cm]
y=5exp(-x)
1
0
0.5
X [kPa]
1
1.5
PROBLEM 1.19
KNOWN: A bulb thermometer is used to measure outside temperature.
FIND: Extraneous variables that might influence thermometer output.
SOLUTION
A thermometer's indicated temperature will be influenced by the temperature of solid objects
to which it is in contact, and radiation exchange with bodies at different temperatures
(including the sky or sun, buildings, people and ground) within its line of sight. Hence,
location should be carefully selected and even randomized. We know that a bulb
thermometer does not respond quickly to temperature changes, so that a sufficient period of
time needs to be allowed for the instrument to adjust to new temperatures. By replication of
the measurement, effects due to instrument hysteresis and instrument and procedural
repeatability can be randomized.
Because of limited resolution in such an instrument, different competent temperature
observers might record different indicated temperatures even if the instrument output were
fixed. Either observers should be randomized or, if not, the test replicated. It is interesting to
note that such a randomization will bring about a predictable scatter in recorded data of about
½ the resolution of the instrument scale.
PROBLEM 1.20
KNOWN:
Input voltage, (Ei) and Load (τL) can be controlled and varied.
Efficiency (η), Winding temperature (Tw), and Current (I) are measured.
FIND: Specify the dependent, independent in the test and suggest any extraneous variables.
SOLUTION
The measured variables are the dependent variables in the test and they depend on the
independent variables of input voltage and load. Several influencing extraneous variables
include: ambient temperature (Ta) and relative humidity R; Line voltage fluctuations (e); and
each of the individual measuring instruments (mi). The variation of the independent variables
should be performed separately maintaining one independent variable fixed while the other is
systematically varied over the test range. A random test procedure for the independent
variable will randomize the effects of Ta, R and e. Replication methods using different test
instruments would be one way to randomize the effects of the mi; alternatively, calibration of
all measuring instruments would provide a good degree of control over these variables.
η= η(EI, τL; Ta, R, e, mi)
Tw = Tw(Ei, τL ;Ta, R, e, mi)
I = I(Ei, τL ; Ta, R, e, mi)
PROBLEM 1.21
KNOWN:
Specifications Table 1.1
Nominal pressure of 500 cm H2O to be measured.
Ambient temperature drift between 18 to 25 oC
FIND: Magnitude of each elemental error listed.
SOLUTION
Based on the specifications:
ri = 1000 cm H2O
ro = 5 V
Hence, K = 5 V/ 1000 cm H2O = 5 mV/cm H2O. This gives a nominal output at 500 cm H2O
input of 2.5 V. This assumes that the input/output relation is linear over range but we are told
that it is linear to within some linearity error.
linearity error = eL = (±0.005) (1000 cm H2O)
= ± 5 cm H2O
= ± 0.025 V
hysteresis error = eh = (±0.0015)(1000 cm H2O)
= ±1.5 cm H2O
= ±0.0075 V
sensitivity error = eK = (±0.0025)(500 cm H2O)
= ± 0.75 cm H2O = ± 0.00375 V
thermal sensitivity error = (±0.0002)(7oC)(500 cm H2O)
= ±0.7 cm H2O
= ± 0.0035 V
thermal drift error = (0.0002)(7oC)(1000 cm H2O)
= 1.4 cm H2O
= 0.007 V
overall instrument error = (52+1.52+0.752+0.72+1.42)1/2 = 5.501 cm H2O
PROBLEM 1.22
KNOWN: FSO = 1000 N
FIND: ec
SOLUTION
From the given specifications, the elemental errors are estimated by:
eL = 0.001 x 1000N = 1N
eH = 0.001 x 1000N = 1N
eK = 0.0015 x 1000N = 1.5N
ez = 0.002 x 1000N = 2N
The overall instrument error is estimated as:
ec = (12 + 12 + 1.52 + 22)1/2 = 2.9N
COMMENT
This root-sum-square (RSS) method provides a "probable" estimate (i.e. the most likely
estimate) of the instrument error possible in any given measurement. "Possible" is a big word
here as error values will most likely change between measurements.
PROBLEM 1.23
SOLUTION
Repetition through repeated measurements made under a fixed set of operating conditions
provides a measure of the time (or spatial) variation of a measured variable.
Replication through the duplication of tests conducted under similar operating conditions
provides a measure of the effect of control of the operating conditions on the measured
variable.
Repetition refers to repeating the measurement during a test.
Replication refers to repeating the test (to repeat the measurements).
PROBLEM 1.24
SOLUTION
Replication is used to assess the ability to control any aspect of a test or its operating
condition. Repeat the test resetting the operating conditions to their original set points.
PROBLEM 1.25
SOLUTION
Randomization is used to break-up the effects of interference from either continuous or
discrete extraneous (i.e. uncontrolled) variables.
PROBLEM 1.26
SOLUTION
This problem does not have a unique solution. We suggest that the instructor use this
problem as a basis for an in-class or small group discussion.
PROBLEM 1.27
FIND: Test matrix to correlate thermostat setting with average room setting
SOLUTION
Although there is no single test matrix, one method of solution follows.
Assume that average room temperature, T, is a function of actual thermostat setting, spatial
distribution of temperature, temporal temperature distribution, and thermostat location. We
might imagine that for a controlled (fixed) thermostat location, a direct correlation between
setting and T could be achieved. However, factors could influence the temperature measured
by the thermostat such as sunlight directly hitting the thermostat or the wall on which it is
attached or a location directly exposed to furnace forced convection, a condition aggrevated
by air conditioners or heat pumps in which delivered air temperature is a strong function of
outside temperature. Assume a proper location is selected and controlled.
Further, the average room temperature must be defined because local room temperature will
vary will position within the room and with time. For the test matrix, the room should be
divided into equal areas with temperature sensing devices placed at the center of each area.
The output from each sensor will be averaged over a time period that is long compared to the
typical furnace on/off cycle.
Select four temperature sensors: A, B, C, D. Select four thermostat settings: s1, s2, s3, s4,
where s1 < s2 < s3 < s4. Temperatures are to be measured under each setting after the room
has adjusted to the new setting. One matrix might be:
BLOCK
1 s1: A, B, C, D
2 s4: A, B, C, D
3 s3: A, B, C, D
4 s2: A, B, C, D
Note that the order of set temperature has been shuffled to attempt to randomize the test
matrix (hysteresis is a common problem in thermostats). The four blocks will yield the
average temperatures, T1, T4, T3, T2. The data can be presented in a form of T versus s.
PROBLEM 1.28
FIND: Test matrix to evaluate fuel efficiency of a production model of automobile
ASSUMPTIONS: Automobile model design is fixed (i.e. neglect options). Require
representative estimate of efficiency.
SOLUTION
Although there is no single test matrix, one method of solution follows. Many variables can
affect auto model efficiency: e.g. individual car, driver, terrain, speed, ambient conditions,
engine model, fuel, tires, options. Whether these are treated as controlled variables or as
extraneous variables depends on the test matrix. Suppose we "control" the options, fuel, tires,
and engine model, that is fix these for the test duration. Furthermore, we can fix the terrain
and the ambient conditions by using a mechanical chassis dynamometer (a device which
drives the wheels with a prescribed mechanical load) in an enclosed, controlled environment.
In fact, such a machine and its test conditions have been specified within the U.S.A. by
government test standards. By programming the dynamometer to start, accelerate and stop
using a preprogrammed routine, we can eliminate the effects of different drivers on different
cars. However, this test will fail to randomize the effects of different drivers and terrain as
noted in the government statement "... these figures may vary depending on how and where
you drive ... ." This leaves the car itself and the test speed as independent variables, xa and xb,
respectively. We defer considering the effects of the instruments and methods used to
compute fuel efficiency until a later chapter, but assume here that this can be done with
sufficient accuracy.
With this in mind, we could choose three representative cars and three speeds with the test
matrix:
BLOCK
1 xa1: xb1, xb2, xb3
2 xa2: xb1, xb2, xb3
3 xa3: xb1, xb2, xb3
Note that since slight differences will exist between cars that can not be controlled, the autos
are treated as extraneous variables. This matrix randomizes the effects of differences between
cars at three different speeds and yields a curve for fuel efficiency versus speed.
As an alternative, we could introduce a driver into the matrix. We could develop a test track
of fixed (controlled) terrain. And we could have three drivers drive three cars at three
different speeds. This introduces the driver as an extraneous variable, noted as A1, A2 and A3
for each driver. Assuming that the tests are run under similar ambient conditions, one test
matrix may be
A1
A2
A3
xa1
xa2
xa3
xb1
xb2
xb3
xb2 xb3
xb3 xb1
xb1 xb2
PROBLEM 1.29
SOLUTION:
Test stand:
Here one would operate the engine under simulated conditions similar to those encountered
at the track – such as anticipated engine RPM and engine load (load: estimated mechanical
loads on the engine due to mechanical losses, tire rolling resistance, aerodynamic resistance,
etc).
Measure:
• fuel and air consumption
• torque and power output
• exhaust gas temperatures to set air:fuel ratio
Track:
Here one would operate the car at conditions similar to those anticipated during the race.
Measure:
• lap time
• wind and temperature conditions (to normalize lap time)
• depending on team other factors can be measured to estimate loads on the car and car
behavior. Clemson Motorsports Engineering has been active in test method
development for professional race teams.
Obvious major differences:
•
•
•
•
Environmental conditions, which effect engine performance, car behavior and tire
behavior.
Engine load on a test stand is well-controlled. On track, the driver does not execute
exactly on each lap, hence varies load such as due to differences in drive path 'line'
and this affects principally aerodynamic loads and tire rolling resistance. Incidentally,
all of these are coupled effects in that a change in one affects the values of the others.
Ram air effect of moving car can be simulated but difficult to get exactly
Each engine is an individual. Even slight differences affect handling and therefore,
how a driver drives the car (thus changing the engine load).
PROBLEM 1.30
KNOWN: Four lathes, 12 machinists are available to produce batches of machine shafts.
FIND: Test matrix to estimate the tolerances held within a batch
SOLUTION
If we assume that batch precision, P, is only a function of lathe and machinist, then
P = f(lathe, machinist)
We can set up a test matrix using all four lathes, L1, L2, L3, L4, and all 12 machinists, A, B,
..., L. The machinists are randomly assigned.
BLOCK
1 L1: A, B, C
2 L2: D, E, F
3 L3: G, H, I
4 L4: J, K, L
Data from each lathe should be indicative of the precision associated with each lathe and the
total ensemble of data indicative of batch precision. However, this test matrix neglects the
effects of shift and day of the week.
One method which treats machinist and lathe as extraneous variables and reduces test size
selects 4 machinists at random. Suppose more than one shaft size is produced at the plant.
We could select 4 shaft diameters, D1, D2, D3, D4 and set up a Latin square matrix:
B
E
G
L
L1
L2
L3
L4
D1
D2
D3
D4
D2
D3
D4
D1
D3
D4
D1
D2
D4
D1
D2
D3
Note that neither matrix includes shift or day of the week effects and these could be
incorporated in an expanded test matrix.
PROBLEM 1.31
SOLUTION
Linearity error
A random static calibration over a specified range will provide the input-output relationship
between y and x (i.e. y = f(x)). A first-order curve fit to this data, for example using a least
squares regression analysis, will provide the fit yL(x). The linearity error is simply the
difference between the measured value of y at any value of x and the value of yL predicted by
the fit at that x.
A manufacturer may wish to keep the linearity error below some target value and, hence,
may limit the recommended operating range for the system for this purpose. In your
experience, you may notice that some systems can be operated outside of their specification
range but be aware their elemental errors may exceed the manufacturer's stated values.
Hysteresis error
A sequential static calibration over a specified range will provide the input-output behavior
between y and x during upscale-only and downscale-only operations. This will tend to
maximize any hysteresis in the system. The hysteresis error is the difference between the
upscale value and the downscale value of y at any given x.
PROBLEM 1.32
KNOWN:
4 brands of tires
8 cars of the same make
FIND: Test matrix to evaluate performance
SOLUTION
Tire performance can mean different things but for passenger tires usually refers to braking
and lateral load adhesion during wet and dry operations. For a given series of performance
tests, performance will depend on tire and car (a tire will perform differently on different
makes of cars). For the same make, subtle differences in production models can affect test
results so we treat the car as an individual and extraneous variable.
We could select 4 cars at random (1,2,3,4) to test four tire brands (A,B,C,D)
1: A, B, C, D
2: A, B, C, D
3: A, B, C, D
4: A, B, C, D
This provides a data pool for evaluating tire performance for a make of car. Note we ignore
the variable of the test driver but this method will incorporate driver variation by testing four
cars. Other strategies could be created.
PROBLEM 1.33
KNOWN: Water at 20oC
Q =f(C,A,dp,ρ)
C = 0.75; D = 1 m
2 < Q < 10 cmm
FIND: Expected calibration curve
SOLUTION
Part of a test matrix is to specify the range of the independent variable and to anticipate the
range resulting in the dependent variable. In this case, the pressure drop will be measured so
that it is the dependent variable during a static calibration. To anticipate the output range of
the calibration then:
Rearranging the known relation,
dp = (Q/CA)2ρ/2
For ρ= 998 kg/m3 (Appendix C), and A = πD2/4, we find:
Q (cmm)
2
5
10
dp (N/m2
1.6
10
40
This is plotted below. It is clear that K will not be a constant as K = f(Q).
P1.33
dp (N/m-m)
50
40
30
20
10
0
0
5
Q (cmm)
10
15
PROBLEM 1.34
SOLUTION
Because Q ∝ dp1/2 and is not linear, the calibration will not be linear. The term ‘linearity’
should not be applied directly. The nonlinear calibration result is just a normal consequence
of the physics.
However, a signal conditioning stage could be inserted within the signal path to produce a
linear output. This is done using logarithmic amplifiers. To illustrate this, plot the calibration
curve in Problem 1.33 on a log-log scale (see C1.6). The result will be a linear curve.
Alternately, you could take the log of each column and plot them on a rectangular scale to get
that same result. A logarithmic amplifier (Chapter 6) performs this same function (as the plot
scale or log key) directly on the signal. A linearity measure can then be extracted with some
meaning.
As flow rate is the variable varied and pressure drop is the variable measured in this
calibration, pressure drop is the dependent variable. The flow rate and the fixed values of
area and density are independent variables.
PROBLEM 1.35
KNOWN:
pistons are sent out for plating
four subcontractors
FIND: Test matrix for quality control
SOLUTION
Consider four subcontractors as A, B, C, D. One approach is to number the pistons and
allocate them to the four subcontractors with subsequent analysis of the plating results. For
example, send 24 pistons each to the four subcontractors and analyze the resulting products
separately. The variation for each subcontractor can be estimated and can be statistically
tested for significant differences.
PROBLEM 1.36
SOLUTION
Controlled variables
A and B (i.e. control the materials of two alloys)
T2 (reference junction temperature)
Independent variable
T1 (measured temperature)
Dependent variable
E (output voltage measured)
A
T1
voltmeter, E
B
T2
PROBLEM 1.37
SOLUTION
Independent variables
micrometer setting (i.e. the applied displacement)
Controlled variable
power supply input
Dependent variable
output voltage measured
Extraneous variables
operator set-up, zeroing of system, and reading of micrometer
ability to set control variables
COMMENT
If you try this you will find that the power supply excitation voltage can have a significant
influence on the results. The ability to provide the exact voltage on replication is important in
obtaining consistent results in many transducers. Even if you use a regulated laboratory
variable power supply, this effect can be seen in your data variation on replication as a
random variation. If you use an unregulated source, be prepared to trace these effects as they
change from hour to hour or from day to day.
Many LVDT units allow for use of dc power, which is then transformed to ac form before
being applied to the coil. It is easiest to see the effect of power setting on the results when
using this type of transducer.
PROBLEM 1.38
SOLUTION
To test for repeatability in the LVDT, we might displace the core to various random values
over a selected range, such as its expected range, and develop a data base. Data scatter about
a curve fit will provide a measure of repeatability for this instrument (methods are discussed
in Chapter 4).
Reproducibility involves re-testing the system at a different facility or equivalent (such as
different instruments and test fixtures). Think of this as a duplication. Even though a similar
procedure and test matrix will be used to test for reproducibility, the duplication involves
different individual instruments and test fixtures. A reproducibility test is a special type of
replication – by using the different facility constraint added. The combined results allow for
interference effects to be randomized.
Bottom Line: The results leading to a reproducibility specification are more representative of
what can be expected by the end user (YOU!).
PROBLEM 1.39
SOLUTION:
(i) Running the car on a chassis dynamometer, which applies a desired load to the wheels
as the car is operated at a desired speed so as to simulate the car being driven, provides a
controlled test environment for estimating fuel consumption. The operating loads form a
'load profile' to simulate the road course.
Allowing a driver to operate a car over a predetermined course provides a realistic
simulation of expected consumption. No matter how well controlled the dynamometer
test, it is not possible to completely recreate the driving situation that a real driver
provides. However, each driver will drive the course a bit differently.
Extraneous variables include: individual entities of driver and of car that affect
consumption in either method; road variations and differences between the test methods;
road or weather conditions (which are both variable) that change the simulation.
(ii) The dynamometer test is well controlled. In the hands of a good test engineer,
valuable information can be ascertained and realistic mileage values obtained. Most
important, testing different car models using a predetermined load profile forms an
excellent basis for comparison between car makes – this is the basis of a 'standarized
test.'.
The variables in a test affect the accuracy of the simulation. Actual values obtained by a
particular driver and car are not tested in a standarized test.
(iii) If the two methods are conducted to represent each other, than these are concomitant
methods. Even if not exact representations, information obtained in one can be used to
get realistic estimates to be expected in the other. For example, a car that gets 10 mpg on
the chassis dynamometer should not be expected to get 20 mpg on the road course.
PROBLEM 1.40
SOLUTION:
This is not an uncommon situation when siblings own similar model cars.
The drivers, the cars, and the routes driven are all extraneous variables in this direct
comparison. Simply put, you and your brother may drive very differently. You both drive
different cars. You likely drive over different routes, maybe very different types of
driving routes. You might live in very different geographic locations (altitude, weather).
The maintenance of the car would play a role, as well.
An arbitrator might suggest that the two of you swap cars for a few weeks to compare. If
the consumption of each car remains the same under different drivers (and associated
different routes, location, etc), then the car is the culprit. If not, then driver and other
variables remain involved.
PROBLEM 1.41
SOLUTION:
The measure of 'diameter' represents an average or nominal value of the object.
Differences along and around the rod affect the value of 'diameter.' Try this with a rod
and a micrometer.
Measurements made at different positions along the rod show 'noise', that is data scatter,
which is handled statistically (that is, we average the values to obtain a single diameter).
Using just a single measurement introduces interference, since that one value may not be
the same as the average value.
Tabulated values of material properties represent average or nominal values. These
should not be confused as being exact values, regardless of the number of decimal places
found in the tables (although the values can be assumed to be reasonably representative to
within a decimal place). Properties of a material will vary with individual specimens – as
such, differences between a nominal value and the actual specimen value will behave as
an interference.
PROBLEM 1.42
SOLUTION
Independent variable:
Applied tensile load
Controlled variable:
Bridge excitation voltage
Dependent variable:
Bridge output voltage (which is related to gauge resistance changes
due to the applied load)
Extraneous variables:
Specimen and ambient temperature will affect gauge resistance
A replication will involve resetting the control variable and specimen and duplicating the
test.
PROBLEM 1.43
SOLUTION
To test repeatability, apply various tensile loads at random over the useful operating range of
the system to build a data base. Be sure to operate within the elastic limit of the specimen.
Direct comparison and data scatter about a curve fit will provide a measure of repeatability
(specific methods to evaluate this are discussed in C4).
Reproducibility involves re-testing the system at a different facility or equivalent (such as
different instruments and test fixtures). Think of this as a duplication. Even though a similar
procedure and test matrix will be used to test for reproducibility, the duplication involves
different individual instruments and test fixtures. Note that the reproducibility test is also a
replication but with the different facility constraint added. The combined results allow for
interference effects to be randomized.
Bottom Line: The results leading to a reproducibility specification are more representative of
what can be expected by the end user (YOU!).
PROBLEM 1.44
This problem is open-ended and does not have a unique solution. This forms a good
opportunity for class discussion.
PROBLEM 1.45
SOLUTION
A car rolling down the hill whose speed is determined by two sensors separated by a distance
s. Car speed could be determined as: speed = (distance traveled)/(elapsed time) = s/(t2 – t1).
The following is a list of the minimum variables that are important in this test:
L: length of car
s: distance between measurements (distance traveled)
θ: angle of inclination
(t2 – t1): elapsed time
where
t1: instance car passes sensor 1
t2: instance car passes sensor 2
Intrinsic assumptions in this test that affect the accuracy of the result:
(1) the speed of the car is actually an average between the speed of the car as it passes sensor
1 and then as it passes sensor 2. The assumption is that any speed change is small in regards
to the measured value. This assumption imposes a systematic error on the measured result.
(2) The length of the car could be a factor if it affects how the sensors are triggered. The car
is assumed to be a point. This assumption may introduce a systematic error into the results.
As for a concomitant approach:
If we consider the gravitational pull as constant (reasonable over a sensible distance), then
the car’s acceleration is simply, a = gsin θ. So its acceleration is easily anticipated and the
ideal velocity at any point along the path can be calculated directly from simple physics. The
actual velocity will be the ideal velocity reduced by resistance effects, including frictional
effects, such as between the car’s wheels and the track and within the wheel axles, and
aerodynamic effects. The actual velocity will be a bit smaller than the ideal velocity, a
consequence of the systematic error in the assumptions. But what it does give us is a value of
comparison for our measurement. If the measured value is markedly different, then we will
know we have some problems in the test.
PROBLEM 1.46
SOLUTION
The power (P) to move a car at any speed (U) equals the aerodynamic drag (D) plus
mechanical drag (M) times the speed plus the parasitic power (Pp) required to turn the
compressor and other mechanical components in the car: i.e.
P = (D+M)U + Pp
With the air conditioning (A/C) off, the parasitic power due to the compressor goes down but
because the windows would then be rolled open, the aerodynamic drag goes up. The
aerodynamic drag increases with speed while the compressor power remains fairly constant
with speed.
To test this question, you might develop a test plan as follows:
Operate the car at several fixed, but well separated, speeds U1, U2, U3 in each of two
configurations, A and B. Configuration A uses the compressor and all windows are rolled up
closed. Configuration B turns the compressor off but driver window is rolled down (open).
Obviously, there can be alternate configurations by rolling down differing windows, but the
idea is the same.
A: U1, U2, U3
B: U1, U2, U3
Concomitant approach: An analytical approach to this problem would tradeoff the power
required to operate the vehicle at different speeds under the two configurations based on
some reasonable published or handbook values (for example, most modern full-sized sedans
have a drag coefficient of about 0.33 based on a frontal area of about 2.1 m2 (these exact
values are for a Toyota Camry) for windows closed, increasing to 0.36 with driver window
open) and maybe about 3HP to run the compressor. But you might research these numbers.
PROBLEM 1.49
This problem is open-ended and does not have a unique solution. Most of these codes can be
found in a library with a quality engineering section or at the appropriate website for the
professional group cited. The results from these searches form a good opportunity for class
discussion.
PROBLEM 1.50
SOLUTION
Transform each relation into the linear form Y = a1X + a0
(a) KNOWN: y = bx m
This function can be rearranged as
m
log
=
y log bx
=
log b + m log x
so if we let
Y= a1 + a 0 X
then
X = log x ; Y = log y ; ao = log b ; a1 = m
(b) KNOWN: y = bemx
identity: ln x = 2.3 log x
This function can be rearranged as
ln y =
ln b + ln emx =
ln b + mx
so if we let
Y= a1 + a 0 X
then
X = x ; a1 = m ; ao = ln b ; Y = ln y
(c) KNOWN: y= b + cm x
This function can be rearranged as
y−b =
cm x
or
log (y-b) = log c + m log x
so if we let
Y= a1 + a 0 X
then
X = log x ; a1 = m ; ao = log c ; Y = log (y-b)
PROBLEM 2.1
FIND: Define signal and provide examples of static and dynamic input signals to
measurement systems.
SOLUTION: A signal is information in motion from one place to another, such as
between stages of a measurement system. Signals have a variety of forms, including
electrical and mechanical.
Examples of static signals are:
1. weight, such as weighing merchandise, etc.
2. body temperature, over the time period of interest
3. length or height, such as the length of a board or a person's height
Examples of dynamic signals:
1. input to an automobile speed control
2. input to a stereo amplifier from a component such as a CD player
3. output signal to a printer from a computer
PROBLEM 2.2
FIND: List the important characteristics of signals and define each.
SOLUTION:
1. Magnitude - generally refers to the maximum value of a signal
2. Range - difference between maximum and minimum values of a signal
3. Amplitude - indicative of signal fluctuations relative to the mean
4. Frequency - describes the time variation of a signal
5. Dynamic - signal is time varying
6. Static - signal does not change over the time period of interest
7. Deterministic - signal can be described by an equation
(other than a Fourier series or integral approximation)
8. Non-deterministic - describes a signal which has no discernible pattern of repetition
andcannot be described by a simple equation.
COMMENT: A random signal, or stochastic noise, represents a truly non-deterministic
signal. However, chaotic systems produce signals that appear random, but are truly
deterministic. An example would be the velocity in a turbulent fluid flow, that may appear
random, but is actually governed by the Navier-Stokes equations.
PROBLEM 2.3
KNOWN:
y (t ) = 30 + 2 cos 6πt
FIND: y and yrms for the time periods t1 to t2 listed below
a) 0 to 0.1 s
b) 0.4 to 0.5 s
c) 0 to 1/3 s
d) 0 to 20 s
SOLUTION:
For the function y(t)
y=
t2
1
y (t ) dt
∫
t
t2 − t1 1
and
yrms =
t2
1
2
y (t ) ] dt
[
∫
t
t2 − t1 1
Thus in general,
y=
t2
1
( 30 + 2cos 6π t ) dt
t2 − t1 ∫t1
t2

1 
2
=
sin 6π t 
30t +
t2 − t1 
6π

t1 
=
1 
2
30 ( t2 − t1 ) +
( sin 6π t2 − sin 6π t1 ) 

t2 − t1 
6π

and
yrms
 1
=
 t2 − t1
 1
=
 t2 − t1

(30 + 2 cos 6πt ) dt 
t1

∫
t2
1
2
2
120

 1 sin 6πt cos 6πt + 1 t   
900
6
4
sin
π
+
+
t
t


 12π
6π
2   t1 
t2
1
2
The resulting values are
a) y 31.01
=
=
yrms 31.02
b) y 28.99
=
=
yrms 29.00
c) y 30
=
=
yrms 30.03
d) y 30
=
=
yrms 30.03
COMMENT: The average and rms values for the time period 0 to 20 seconds
represents the long-term average behavior of the signal. The values which result in parts
a) and b) are accurate over the specified time periods, and for a measured signal may have
specific significance. If we examined the period 0 to 1/3, it would represent one complete
cycle of the simple periodic signal and results in average and rms values which accurately
represent the long-term behavior of the signal.
PROBLEM 2.4
KNOWN: Discrete sampled data, corresponding to measurement every 0.4 seconds.
FIND: The mean and rms values of the measured data.
SOLUTION:
The mean value for y1 is 0 and for y2 is also 0.
However, the rms value of y1 is 13.49 and for y2 is 17.53.
COMMENT: The mean value contains no information concerning the time varying
nature of a signal; both these signals have an average value of 0. But the differences in
the signals are made apparent when the rms value is examined.
PROBLEM 2.5
KNOWN: The effect of a moving average signal processing technique is to be
determined for the signal in Figure 2.21 and =
y (t ) sin 5t + cos11t
FIND: Discuss Figure 2.22 and plot the signal resulting from applying a moving average
to y(t).
ASSUMPTIONS: The signal y(t) may be represented by making a discrete
representation with δt = 0.05.
SOLUTION:
a) The signal in Figure 2.22 clearly has a reduced level of high frequency content. In
essence, this emphasizes longer term variations, while removing short-term fluctuations.
It is clear that the peak-to-peak value in the original signal is significantly higher than in
the signal that has been averaged.
b) The figures below show in the effect of applying a moving average to y(t).
Signal y(t) = sin 5t + cos 11t
2.5
2
1.5
1
y
0.5
0
-0.5
-1
-1.5
-2
-2.5
0
1
2
3
t
4
5
6
Four Point Moving Average
2.5
2
1.5
1
0
-0.5
-1
-1.5
-2
-2.5
0
1
2
3
4
5
6
t
30 point moving average
1.2
1
0.8
0.6
0.4
0.2
y
y
0.5
0
-0.2
-0.4
-0.6
-0.8
-1
0
1
2
3
4
t
5
6
7
PROBLEM 2.6
KNOWN: A spring-mass system, with
m = 0.5 kg
T = 2.7 s
FIND: Spring constant, k, and natural frequency ω
SOLUTION:
Since
ω=
k
m
(as shown in association with equation 2.7)
and
2π 1
= = 2.7 s
f
ω
ω = 2.33 rad/s
T=
The natural frequency is then found as ω = 2.33 rad/s
And
=
=
ω 2.33
k
0.5 kg
k = 2.71 N/m (kg/sec 2 )
PROBLEM 2.7
KNOWN: A spring-mass system having
m = 1 kg
k = 5000 N/cm
FIND: The natural frequency in rad/sec (ω) and Hz (f ).
SOLUTION:
The natural frequency may be determined,
ω=
k
=
m
5000
and
f=
ω
= 112.5 Hz
2π
N
cm
100
cm
m = 707.1 rad / s
1 kg
PROBLEM 2.8
KNOWN: Functions:
2π t
5
b) 5cos 20t
a) sin
c) sin 3nπt for n = 1 to ∞
FIND: The period, frequency in Hz, and circular frequency in rad/s.
SOLUTION:
a) ω = 2π/5 rad/s
b) ω = 20 rad/s
c) ω = 3nπ rad/s
f = 0.2 Hz
f = 3.18 Hz
f = 3n/2 Hz
T =5s
T = 0.31 s
T = 2/(3n) s
PROBLEM 2.9
KNOWN:=
y (t ) 5sin 4t + 3cos 4t
FIND: Equivalent expression containing a cosine term only
SOLUTION:
y = C cos( ωt − φ )
From Equations 2.10 and 2.11
φ = tan −1
B
A
and with
A cosωt + B sin ωt =
A 2 + B 2 cos( ωt − φ )
we find
C=
A2 + B 2 =
52 + 32 = 5.83
−1 5
1.03 rad
=
=
φ tan
3
and
=
y ( t ) 5.83cos ( 4t − 1.03)
PROBLEM 2.10
KNOWN:
=
y (t ) 4sin 2π t + 15cos 2π t
FIND:
a) Equivalent expression containing a cosine term only
b) Equivalent expression containing a sine term only
SOLUTION:
From Equations 2.10 and 2.11
B
A
y = C cos(ωt − φ )
φ = tan −1
y = C sin(ωt + φ * )
φ * = tan −1
A
B
and with
A cos ωt + B sin ωt =
A 2 + B 2 cos(ωt − φ )
A cos ωt + B sin ωt =
A 2 + B 2 sin(ωt + φ * )
we find
C=
A2 + B 2 =
152 + 42 = 15.52
−1 4
0.26 rad
=
φ tan
=
15
−1 15
1.31 rad
=
φ * tan
=
4
and
=
y 15.52 cos ( 2π t − 0.26 ) Answer (a)
=
y 15.52sin ( 2π t + 1.31) Answer (b)
PROBLEM 2.11
∞
KNOWN: y(t ) =
∑ 2π6n sin nπt + 4π6n cos nπt
n =1
FIND:
a) Equivalent expression containing a cosine term only
SOLUTION:
From Equation 2.19
∞
y=
C cos(ωt − φ )
∑
n=1
n
φ n = tan −1
n
Bn
An
with
Cn = An2 + Bn2
and with
2πn 
4πn 
Cn = 
 + 
 =
 6 
 6 
2
2
5
πn
9
and φ n = tan
−1
(2πn 6) = tan
(4πn 6)
we find
∞
y( t ) =
∑
n =1
5
πn cos(nπt − 0.46)
9
−1
0.5 = 0.46 rad
PROBLEM 2.12
KNOWN: T is a period of y(x)
FIND: Show that nT for n=2,3,... is a period of y(x)
SOLUTION: Since T is a period of y(x)
y(x + T) = y(x)
Letting x1 = x + T yields
y(x1) = y(x) = y(x + T)
But since T is a period of y(x)
y(x + 2T) = y(x1 + T) = y(x)
By analogy then
y(x + nT) = y(x)
PROBLEM 2.13
KNOWN:
∞
y (t ) = ∑
3n
5n
sin nt + cos nt
3
n =1 2
FIND:
a) fundamental frequency and period
b) cosine series
SOLUTION:
a) The fundamental frequency corresponds to n = 1, so ω = 1 rad/s;
b) From equation 2.19

 2nπt
y (t ) = Ao + ∑ Cn cos
− φn 

 T
n =1
Bn
Cn = An2 + Bn2
tan φ n =
An
∞
For this Fourier series
181 2
 3n   5n 
Cn =   +   =
n
2 3
36
2
2
9
 ⇒ φ = 0.7328
 10 
φ n = tan −1 
Thus the series may be written
=
y (t )
∞
∑
n =1
181 2
n cos ( nt − 0.7328 )
36
T = 2π
PROBLEM 2.14
KNOWN:
∞
2nπ
nπ
120nπ
nπ
y (t ) =
4+∑
cos
t+
sin
t
4
30
4
n =1 10
FIND:
a) ω1 and f1
b) T1
c)
∞
 2nπ t

y (t ) =
Ao + ∑ Cn sin 
+φ* 
 T

n =1
SOLUTION:
a) When n ==
1, ω1 π=
f 1
4, 1
8
b) T1 = 8 sec
c) From Eq. (2.21)
An
Cn =
An2 + Bn2 and tan φ * =
Bn
 2nπ   120nπ 
Cn = 
 +
 =4nπ
 10   30 
2
φn* tan −1
=
2nπ 10 )
(=
(120nπ 30 )
2
−1  1 
tan
=
  0.05 rad
 20 
and the Fourier sine series
∞
 nπ t

y (t ) =
+ 0.05 
4 + ∑ 4nπ sin 
 4

n =1
PROBLEM 2.15
KNOWN:
(t ) 0 for − π ≤ t ≤ 0
y=
y (t ) =−1 for 0 ≤ t ≤ π
=
y (t ) 1 for π
FIND:
2
2
≤t ≤π
Fourier series for y(t)
SOLUTION: Since the function is neither even nor odd, the Fourier series will contain
both sine and cosine terms. The coefficients are found as
1 −T 2
1 π
(
)
=
y
t
dt
y (t )dt
2π ∫−π
T ∫−T 2
π
π
1  0

2
0
=
+
−1dt + ∫π 1dt 
dt
∫
∫

0
2π  −π
2

1  −π
=
− 0 + π − π  = 0 ∴ Ao = 0
2
2 
2π 
=
Ao
(
) (
)
Note: Since the contribution from −π to 0 is identically zero, it will be omitted.
π
2  π
2nπ t
2nπ t 
An =  ∫ 2 −1cos
dt
+ ∫π 1cos
T
T 
2π  0
2
π
π
1   −1
 2 1
 
sin nt  +  sin nt  
=

π   n
0  n
π 2 

−1
 nπ 
2sin 
=

πn
 2 
π
2  π
2π nt
2π nt 
Bn =  ∫ 2 −1sin
dt + ∫π 1sin
dt 
0
2π 
T
T
2

π
1
1
π
=
 − cos ( 0 ) − cos ( nπ ) 
[ cos nt ]0 2 + [ − cos nt ]π 2=
nπ
nπ 
Noting that An is zero for n even, and Bn is zero for n odd, the resulting Fourier series is
{
}
2
1
1
1
1
1
1

y (t ) =−
cos
t
sin
2
t
cos
3
t
sin
4
t
cos
5
t
sin
6
t
cos 7t − ...
−
+
−
−
−
+

2
3
4
5
6
7
π

Problem 2.15
1.5
First Partial Sum
Third Partial Sum
Fourteenth Partial Sum
1
0.5
t
0
-0.5
-1
-1.5
-4
-3
-2
-1
0
y(t)
1
2
3
4
PROBLEM 2.16
KNOWN: y(t) = t for −5 < t < 5
FIND: Fourier series for the function y(t).
ASSUMPTIONS: An odd periodic extension is assumed.
SOLUTION:
The function is approximated as shown below
5
3
1
y(t) -15
-5 -1
5
15
-3
-5
t
Since the function is odd, the Fourier series will contain only sine terms
∞
y (t ) = ∑ Bn sin
n =1
2nπt
T
where, from (2.17)
T
2
Bn =
T
∫
Bn =
2
10
2
y(t )sin
−T 2
2nπt
dt
T
Thus
which is of the form x sin ax , and
∫
5
−5
t sin
2 nπt
dt
10
2  5 
2nπt 10t
2nπt 
−
cos
Bn =   sin

10  nπ 
10
2nπ
10 
2
5
−5
2

2  5 
 50 
{cos( nπ ) + cos( − nπ ) } 
 {sin( nπ ) − sin( − nπ ) } − 
= 
 2nπ 
10  nπ 

for n even Bn =
for n odd Bn =
−10
nπ
10
nπ
The resulting Fourier series is
y (t ) =
10
π
sin
2πt 10
4πt 10
6πt 10
8πt
−
sin
+
sin
−
sin
+ - ...
10 2π
10 3π
10 4π
10
5
4
3
2
1
y
0
-5
-4
-3
-2
-1
-1
0
-2
-3
-4
-5
t
1
2
3
4
5
PROBLEM 2.17
KNOWN: =
y (t ) t 2 for − π ≤ t ≤ π ; y (t + =
2π ) y (t )
FIND: Fourier series for the function y(t).
SOLUTION:
Since the function y(t) is an even function, the Fourier series will contain only cosine
terms,
∞
y(t ) = Ao +
∑
n =1
2 nπt
An cos
= Ao +
T
∞
∑ A cos nωt
n
n =1
The coefficients are found as
1
Ao =
T
An =
1
π
∫
π
t 2 cos
−π
∫
T
2
y(t )dt =
−T 2
1
2π
∫
π
t 2 dt =
−π
π2
3
2 nπt
dt
2π
π

1  2t cos nt n 2 t 2 − 2
= 
+
sin nt 
3
2
n
π n
 −π
=
1  2π
2π
+
n
cos
cos( − nπ )
π
(
)
2
π  n
n2

for n even An= 4/n2 for n odd An = −4/n2 and the resulting Fourier series is
y( t ) =
π2
1
1
− 4 cos t − cos 2t + cos 3t − +...
3
4
9


a series approximation for π is
y(π ) = π 2 =
or
π2
6
=1+
π2
1
1
− 4 cos π − cos 2π + cos 3π − +...
3
4
9


1 1 1
+ + +...
4 9 16
PROBLEM 2.18
KNOWN:
FIND:
t for 0 < t < 1 
y(t) = 

2 − t for 1 < t < 2 
Fourier series representation of y(t)
ASSUMPTION: Utilize an odd periodic extension of y(t)
SOLUTION:
The function is extended as shown below with a period of 4.
Odd Periodic Extension
1
y
0.5
0
-2
-1
-0.5 0
1
2
-1
t
The Fourier series for an odd function contains only sine terms and can be written
∞
y( t ) =
∑
n =1
2 nπt
Bn sin
=
T
∞
∑ B sin nωt
n
n =1
where
Bn =
∫
T
2
−T
y(t )sin
2
2 nπt
dt
T
For the odd periodic extension of the function y(t) shown above, this integral can be
expressed as the sum of three integrals
Bn =
∫
−1
−(2 + t )sin
−2
2 nπt
dt +
4
∫
1
t sin
−1
2 nπt
dt +
4
∫
2
( 2 − t ) sin
1
2nπt
dt
4
These integrals can be evaluated and simplified to yield the following expression for Bn
nπ
− sin( nπ ) + 2 sin 
 2 
Bn = 4
n 2π 2
Since sin(nπ) is identically zero, and sin(nπ/2) is zero for n even, the Fourier series can be
written
y( t ) =
8  πt 1
3πt 1
5πt 1
7πt
sin − sin
+ sin
− sin
+ −...
2 
2
25
2
49
2
π  2 9

The first four partial sums of this series are shown below
First Four Partial Sums
1
First Partial Sum
Second Partial Sum
Third Partial Sum
Fourth Partial Sum
0.9
0.8
0.7
y
0.6
0.5
0.4
0.3
0.2
0.1
0
0
0.5
1
Time (Sec)
1.5
2
PROBLEM 2.19
KNOWN:
a) sin 10t V
b) 5 + 2 cos 2t m
c) 5t s
d) 2 V
FIND: Classification of signals
SOLUTION:
a)
b)
c)
d)
Dynamic, deterministic, simple periodic waveform
Dynamic, deterministic periodic with a zero offset
Dynamic, deterministic, unbounded as t → ∞
Static, deterministic
PROBLEM 2.20
KNOWN: At time zero (t = 0)
x=0
dx
= 5 cm / s f = 1 Hz
dt
FIND:
a) period, T
b) amplitude, A
c) displacement as a function of time, x(t)
d) maximum speed
SOLUTION:
The position of the particle as a function of time may be expressed
x (t ) = A sin 2πt
so that
dx
= 2 Aπ cos 2πt
dt
dx
Thus, at t = 0
=5
dt
From these expressions we find
a) T = 1 s
b) amplitude, A = 5/2π
c) x (t ) = A sin 2πt
d) maximum speed = 5 cm/s
PROBLEM 2.21
KNOWN:
a) Frequency content
b) Amplitude
c) Magnitude
d) Period
FIND:
Define the terms listed above
SOLUTION:
a) Frequency content - for a complex periodic waveform, refers to the relative amplitude
of the terms which comprise the Fourier series for the signal, or the result of a Fourier
transform.
b) Amplitude - the range of variation of a particular frequency component in a complex
periodic waveform
c) Magnitude - the value of a signal, which may be a function of time
d) Period - the time for a signal to repeat, or the time associated with a particular
frequency component in a complex periodic waveform.
PROBLEM 2.22
KNOWN:
Fourier series for the function y(t) = t in Problem 2.16
y (t ) =
10
π
sin
2π t 10
4π t 10
6π t 10
8π t
sin
sin
sin
−
+
−
+
10 2π
10 3π
10 4π
10
FIND:
Construct an amplitude spectrum plot for this series.
SOLUTION:
Amplitude Spectrum
3.5
Amplitude
3
2.5
2
1.5
1
0.5
0
0
0.1
0.2
f (Hz)
0.3
0.4
PROBLEM 2.23
KNOWN: Fourier series for the function y(t) = t2 in Problem 2.17
y (t ) =
π2
1
1


− 4 cos t − cos 2t + cos 3t − +...


9
4
3
FIND:
Construct an amplitude spectrum plot for this series.
SOLUTION:
The corresponding frequency spectrum is shown below
Amplitude Spectrum
4.5
4
3.5
Amplitude
3
2.5
2
1.5
1
0.5
0
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
f (Hz)
COMMENT: The relative importance of the various terms in the Fourier series as
discerned from the amplitude of each term would aid in specifying the required frequency
response for a measurement system. For example, the term cos 4x has an amplitude of
1/16, which for many purposes may not influence a measurement, and would allow a
measurement system to be selected to measure frequencies up to 0.6 Hz.
PROBLEM 2.24
KNOWN: Signal sources:
a) thermostat on a refrigerator
b) input to a spark plug
c) input to a cruise control
d) a pure musical tone
e) note produced by a guitar string
f) AM and FM radio signals
FIND: Sketch representative signal waveforms.
SOLUTION:
a)
Signal
Time
b)
Signal
Time
c)
Cruise Control
100
90
80
Signal
70
60
50
40
30
20
Time
d)
Signal
Pure Musical Tone
0
1
2
3
4
Time
e)
Time
Guitar Note
Signal
5
6
7
FM Radio Wave
-0.5
-1
-1.5
-2
-2.5
time
386
375
364
353
342
331
320
309
298
287
276
265
254
243
232
221
210
199
188
177
166
155
144
133
122
111
100
89
78
67
56
45
34
23
12
1
amplitude
f)
AM radio wave
2
1.5
1
0.5
0
Series1
PROBLEM 2.25
KNOWN:
=
e(t ) 5sin 31.4t + 2sin 44t Volts
FIND: e(t) as a discrete-time series of N = 128 numbers separated by a time increment of
δt . Find the amplitude-frequency spectrum.
SOLUTION:
With N = 128 and δt = 1/N, the discrete-time series will represent a total time (or series
length) of N δt = 1 sec. The signal to be represented contains two fundamental
frequencies,
f1 = 31.4/2π = 5 Hz and
f2 = 44/2π = 7 Hz
We see that the total time length of the series will represent more than one period of the
signal e(t) and, in fact, will represent 5 periods of the f1 component and 7 periods of the f2
component of this signal. This is important because if we represent the signal by a
discrete-time series that has an exact integer number of the periods of the fundamental
frequencies, then the discrete Fourier series will be exact.
Any DFT or FFT program can be used to solve this problem. Using the
companion software disk, issues associated with sampling continuous signals to create
discrete-time series. The time series and the amplitude spectrum are plotted below.
5sin31.4t + 2 sin44t Volts
8
6
4
e(t)
2
0
-2
-4
-6
-8
0
0.2
0.4
0.6
Time (sec)
0.8
1
1.
Frequency (Hz)
63
61
59
57
55
53
51
49
47
45
43
41
39
37
35
33
31
29
27
25
23
21
19
17
15
13
11
9
7
5
3
1
Signal Amplitude
Amplitude Spectrum
6
5
4
3
2
1
0
PROBLEM 2.27
KNOWN: 3250 µε < ε < 4150 µε , f = 1 Hz
FIND:
a) average value
b) amplitude and frequency when the output is expressed as a simple
periodic function
c) express the signal, y(t), as a Fourier series.
SOLUTION:
a) Ao
=
b)
=
+ 4150 ) 2 3700 µε
average value = ( 3250
C1 =
450 µε
( 4150 − 3250 ) 2 =
f1 = 1 Hz
(
c) y (t ) =
3700 + 450sin 2π t ± π
2
)
µε
PROBLEM 2.28
KNOWN: A force input signal varies between 100 and 170 N (100 < F < 170 N) at a
frequency of ω = 10 rad/s.
FIND: Signal average value, amplitude and frequency. Express the signal, y(t), as a
Fourier series.
SOLUTION:
The signal characteristics may be determined by writing the signal as
y(t) = 135 + 35 sin 10t [N]
a) Ao = Average value = (170 + 100)/2 = 135 N
b) C1 = (170 − 100)/2 = 135 N; f1 = ω/2π = 10/2π = 1.59 Hz
c) y(t) = 135 + 35 sin (10t ± π/2)
PROBLEM 2.29
KNOWN: A periodic displacement varies between 2 and 5 mm (2 < x < 5) at a
frequency of f = 100 Hz.
FIND: Express the signal as a Fourier series and plot the signal in the time domain, and
construct an amplitude spectrum plot.
SOLUTION:
Noting that
=
Ao Average value
=
= ( 2 + 5) 2
3.5 mm
C1 =
( 5 − 2 ) 2 =1.5 mm
f1 = 100 Hz
The displacement may be expressed
(
y (t ) =
3.5 + 1.5sin 200π t ± π
2
) mm
The resulting time domain behavior is a simple periodic function; an amplitude spectrum
plot should show a value of 3.5 mm at zero frequency and a value of 1.5 mm at a
frequency of 100 Hz.
PROBLEM 2.30
KNOWN: Wall pressure is measured in the upward flow of water and air. The flow is
in the slug flow regime, with slugs of liquid and large gas bubbles alternating in the flow.
Pressure measurements were acquired at a sample frequency of 300 Hz, and the average
flow velocity is 1 m/sec.
FIND: Construct an amplitude spectrum for the signal, and determine the length of the
repeating bubble/slug flow pattern.
SOLUTION:
The figure below shows the amplitude spectrum for the measured data. There is clearly a
dominant frequency at 0.73 Hz. Then with an average flow velocity of 1 m/sec, the length
is determined as
=
L
1 m/sec
= 1.37 m
0.73 Hz
Slug Flow Data
2500
2000
Amplitude
1500
1000
500
0
0
2
4
6
8
10
Frequency
12
14
16
18
20
PROBLEM 2.31
KNOWN:
Signals,
a) Clock face having hands
b) Morse code
c) Musical score
d) Flashing neon sign
e) Telephone conversation
f) Fax transmission
FIND: Classify signals as completely as possible
SOLUTION:
a) Analog, time-dependent, deterministic, periodic, steady-state
b) Digital, time-dependent, nondeterministic
c) Digital, time-dependent, nondeterministic
d) Digital, time-dependent, deterministic
e) Analog or digital, time-dependent, nondeterministic
f) Digital, time-dependent, nondeterministic
PROBLEM 2.32
KNOWN: Amplitude and phase spectrum for {y(rδt)} from Figure 2.26
FIND: {y(rδt)}, δ f , δ t
SOLUTION:
By inspection of Figure 2.27:
=
C1 5=
V C2 0=
V C3 3=
V C4 0=
V C5 1 V
=
f1 1=
Hz f 2 2 =
Hz f3 3 =
Hz f 4 4=
Hz f5 5 Hz
=
rad φ2 0=
rad φ3 0.2=
rad φ4 0=
rad φ5 0.1 rad
φ1 0=
and δ f = 1 Hz .
The signal can be reconstructed from the above information, as
y (t ) 5sin ( 2π t ) + 3sin ( 6π t + 0.2 ) + sin (10π t + 0.1)
=
The exact phase of the signal relative to t = 0 is not known, so y(t) is ambiguous within
± π 2 in terms of its overall phase.
A DFT returns N/2 values. Therefore 5 spectral values implies that N = 10. Then
=
δ f 1 N=
δ t 1=
Hz 1 10 δ t giving=
δ t 0.1 sec or=
f s 10 Hz
Alternatively, by inspection of the plots
=
f N f=
=
f s 10 Hz or=
δ t 0.1 sec
5 Hz giving
s 2
PROBLEM 2.33
KNOWN:
( 4C T ) t + C
y (t ) = ( − 4C T ) t + C
=
y (t )
−T 2 ≤ t ≤ 0
0≤t ≤T 2
FIND: Show that the signal y(t) can be represented by the Fourier series
∞
y(t ) = Ao +
∑
n =1
4C(1 − cos nπ )
2nπt
cos
2
T
( πn )
SOLUTION:
a) Since the function y(t) is an even function, the Fourier series will contain only cosine
terms,
∞
y(t ) = Ao +
∑
n =1
2 nπt
An cos
= Ao +
T
∞
∑ A cos nωt
n
n =1
The value of Ao is determined from Equation (2.17)
T
1 2
Ao =
y(t )dt
T −T 2
∫
Ao =
∫
0
−T
2
 4Ct + C dt +
 T

∫
T
2
0
 −4Ct + C dt


 T

integrating yields a value of zero for Ao

 −2Ct 2
 2 
1  2Ct 2
Ao = 
+ Ct  + 
+ Ct   = 0
T  T
−T 2  T
 0 

T
0
Then to determine An
2 
An = 
T 
∫
0
−T
2
 4Ct + C cos 2 nπt dt +


 T

T
∫
T
0
2
 −4Ct + C cos 2 nπt dt 



 T

T

C( −2 + 2 cos( nπ ) + nπ sin( nπ ))
( nπ ) 2
Since sin(nπ) = 0, then the Fourier series is
∞
4C(1 − cos nπ )
2 nπt
cos
y( t ) =
2
T
( πn )
n =1
An = −2
∑
The values of An are zero for n even, and the first three nonzero terms of the Fourier
series are
8C
2πt
8C
6πt
8C
10πt
cos
+
cos
+
cos
2
2
2
T (3π )
T (5π )
T
(π )
The first term represents the fundamental frequency.
y
Separate Representation of First Three Terms
1
0.8
0.6
0.4
0.2
0
-0.2 0
-0.4
-0.6
-0.8
-1
0.5
1
1.5
2
t
Third Partial Sum
1
y
0.5
0
0
0.5
1
-0.5
-1
t
1.5
2
PROBLEM 2.33
KNOWN: Figure 2.16 illustrates the nature of spectral distribution or frequency
distribution on a signal.
FIND: Discuss the effects of low amplitude high frequency noise on signals.
SOLUTION:
Assume that Figure 2.16a represents a signal, and that Figures 2.16 b-d represent the
effects of noise superimposed on the signal. Several aspects of the effects of noise are
apparent. The waveform can be altered significantly by the presence of noise, particularly
if rates of change of the signal are important for specific purposes such as control.
Generally, high frequency, low amplitude noise will not influence a mean value, and most
of the signal statistics are not affected when calculated for a sufficiently long signal.
PROBLEM 3.1
KNOWN:
K = 2 V/kg
F(t) = constant = A
Possible range of A: 1 kg to 10 kg
FIND:
y(t)
SOLUTION
We will model the input as a static value and interpret the static output that results. To do
this, this system is modeled as a zero order equation.
y(t) = KF(t)
where F(t) is constant for all time; so y(t) is constant
At the low end of the range, F(t) = A = 1 kg, then
y = KF(t) = (2 V/kg)(1 kg) = 2 V
At the high end of the range, F(t) = A = 10 kg, then
y = KF(t) = (2 V/kg)(10 kg) = 20 V
Hence, the output will range from 2 V to 20 V depending on the applied static input value.
Clearly, the model shows that if K were to be increased, the static output y would be
increased. Here K is a constant, meaning that the relationship between the applied input and
the resulting output is constant. The calibration curve must be a linear one. Notice how K,
through its value, takes care of the transfer in the units between input F and output y.
COMMENT
Because we have modeled this system as a zero order responding system,
we have eliminated any accommodation for a transient response in the system model
solution. The forcing function (i.e., input signal) is constant (i.e., static) for all time. So in the
transient sense, this solution for y is valid only under static conditions. However, it is correct
in its prediction in the steady output value following any change in input value.
PROBLEM 3.2
KNOWN: System model
FIND: 75%, 90% and 95% response times
ASSUMPTIONS: Unless noted otherwise, all initial conditions are zero.
SOLUTION
We seek the rise time to 75%, 90% and 95% response. For a first order system, the percent
response time is found from the time response of the system to a step change in input. The
error fraction for such an input is given by
C(t) = e-t/τ
from which the percent response at time t is found by
% response = (1 - Γ(t)) x 100
= (1 - e-t/τ) x 100
from which t is computed directly. Alternatively, Figure 3.7 could be used.
For a second order system, the system response depends on the damping ratio and natural
•
a )0.4 T + T = 4U (t )
frequency of the system and can be established from either (3.15) or Figure 3.14.
By direct comparison to the general model of a first order system, τ = 0.4 s. Hence,
75% = (1 - e-t/0.4) x 100
or t75 = 0.55 s
90% = (1 - e-t/0.4) x 100
or t90 = 0.92 s
95% = (1 - e-t/0.4) x 100
or t95 = 1.2 s
Alternatively, Figure 3.7 could be used.
Problem 3.2 continued
••
•
b) y + 2 y + 4 y = U ( t )
By direct comparison to equations (3.12) and (3.13),
ζ = 0.5 and ωn = 2 rad/s
Then using Figure 3.14 as a guide, for a response of
75%:
ωn t ~ 1.75 so that t75 ~ 0.9 s
90%:
ωn t ~ 2
95%:
ωn t ~2.4 so that t95 ~ 1.2 s
••
so that t90 ~ 1.0 s
•
c )2 P + 8 P + 8P = 2U (t )
By direct comparison to equations (3.12) and (3.13),
ζ = 1.0 and ωn = 2 rad/s
Then using Figure 3.14 as a guide, for a response of
75%:
ωn t ~ 2.6 so that t75 ~ 1.3 s
90%:
ωn t ~3.7 so that t90 ~ 1.9 s
95%:
ωn t ~ 4.6 so that t95 ~ 2.3 s
•
d )5 y + 5 y = U (t )
By direct comparison to the general model of a first order system,
τ = 1 s and K = 0.2 units/unit.
75% = (1 - e-t/τ) x 100
or t75 = 1.4 s
90% = (1 - e-t/τ) x 100
or t90 = 2.3 s
95% = (1 - e-t/τ) x 100
or t95 = 3.0 s
PROBLEM 3.3
KNOWN: let X = % vapor
X
y [units]
0
50
100
80
40
0
FIND: K
SOLUTION
The data fit the linear curve,
y = -0.8x + 80
The static sensitivity is defined as
K = dy/dx x = -0.80 units/% vapor
where K is independent of input x.
COMMENT
Because this system's calibration curve is linear, the static sensitivity remains constant over
the input range.
Be certain to always provide units for all answers; magnitudes alone are not sufficient.
PROBLEM 3.4
KNOWN: System model equation
K = 1 unit/unit
F(t) = 100U(t)
y(0) = 75 units
FIND: y(t)
SOLUTION
(a) The solution to the system model was shown to be given by the general form,
y(t) = y∞ + (y(0) - y∞ )e-t/τ
where here,
y(0) = 75 units
y∞ = 100 units
τ = 0.5 s (determined from the system equation)
then,
y(t) = 100 + (75 - 100) e-t/0.5 units
Alternatively, by direct solution and with y(0) = 75 units,
•
0.5 y + y =
100U (t )
for t ≥ 0+
y(t) = yh + yp
= Ce-t/τ + B
By substitution, B = 100. Then, if y(0) = 75 and τ = 0.5, C = -25.
y(t) = 100 - 25 e-t/0.5
Temperature ( C)
(b) The input signal and output signal are shown below.
P3.4
125
y(t)
F(t)
100
75
50
0
2
t(sec)
4
6
PROBLEM 3.5
KNOWN:
Thermometer similar to Example 3.3
First order system model
K = 1 oF/oF
τ = 30 s
F(t) = AU(t) = (120 - 32 oF)U(t)
T(0) = 32 oF
FIND: T(t); 90% rise time
SOLUTION
(a) From Example 3.3,
•
mc T + hA[T (t ) − T (0)] = hA[T∞ − T (0)]U (t )
or
•
τ T + T = 120 o F
for t ≥0+ with T(0) = 32 oF. Subbing yields
T(t) = 120 - 88e-t/30 for t ≥0+
This response is plotted below.
P3.5
T(t) [ F]
140
120
100
80
60
40
20
0
0
100
t [sec]
200
300
(b) To find the 90% response or rise time,
%response = 1 - Γ(t) = 1 – e-t/τ
For τ = 30 s and Γ = 0.1 , this yields
t90 = 69 s
PROBLEM 3.6
SOLUTION:
We saw in Example 3.3 that the time constant for a thermal system was dependent on
ambient conditions. Specifically, the heat transfer coefficient is dependent on such
conditions.
If a student were to remove a sensor from hot water and transfer it to cold water by hand, it
would be in motion part of the time. Further, one student may hold the sensor in the cold bath
more steadily than another. Movement will change the heat transfer coefficient on the sensor
by a factor of from 2 to 5 or more. Hence, the variation in time constant noted between
students was simply a lack of control of the heat transfer coefficient – that is, a lack of
control of the test condition.
Their answers (results) are not incorrect, just inconsistent! The results simply show the
effects of a random error, in this case due to variations in the test condition. By proper test
plan design, they can obtain a reasonable result that is bracketed by their test uncertainty.
This uncertainty can be quantified by methods developed in Chapters 4 and 5.
PROBLEM 3.7
KNOWN:
First order instrument
τ = 20 ms
FIND:
Rise time
SOLUTION
The percent response of a first order system subjected to a step input is
%response = 1 - Γ(t) = 1 - e-t/τ
For example, for a 90% rise time, Γ(t) = 0.1. Solving for time,
t90 = 2.3τ
= 46 ms
PROBLEM 3.8
KNOWN: Dynamic calibration using a step input
y∞ - y(0) = 100 units
y(t = 1.2 s) = 80 units
y(0) = 0 units
FIND: τ, y(t = 1.5 s)
SOLUTION
A first order system subjected to a step input can be modeled as
•
τ y + y = KAU(t)
with y(0). The solution is given by (3.5) as
y(t) = KA + (y(0) - y∞ ) e-t/τ
From the information provided, we establish that KA = 100 units. Then,
y(t) = 100 - 100 e-t/τ
Using the information that y(1.2) = 80 units, we solve for τ,
y(1.2) = 80 units = 100 - 100 e-t/τ units
then
τ= 0.75 s.
At time 1.5 s,
y(1.5) = 100 - 100 e-1.5/0.75 units = 86.5 units
PROBLEM 3.9
KNOWN:
First order system
τ = 0.7 s
KF(t) = KAsin 4πt (note here, ω = 2π f where f = 2 Hz)
FIND: δ
SOLUTION
The dynamic error is defined as
δ (ω) = M(ω) - 1
where for a first order system
M(ω)= 1/[1 + (τω)2]½
Now, ω = 2πf so that
M(f) = 1/[1 + (2πf τ)2]½
By direct substitution,
M(f) = 1/[1 + (2.8π)2]½ = 0.11
So that the dynamic error,
δ(2 Hz) = δ(4π rad/s) = - 0.89
COMMENT
This result means that the output amplitude of the 2 Hz signal will be 89% smaller than the
sensed input amplitude KA. That is, the value KA will be attenuated by a factor of 0.89.
Attenuation results with a negative value for δ while a positive value indicates a gain.
PROBLEM 3.10
KNOWN:
First order instrument
τ= 1 s
K = 1 unit/unit
F(t) = 10cos 2.5t = 10sin (2.5t π/2)
y(0) = 0
ysteady(t), β1
FIND:
SOLUTION
For a first order system subjected to a simple periodic waveform input signal, the output
response has the form
y(t) = ce-t/τ+ M(ω)KAsin (2.5t + π/2 + φ(ω))
The steady response is given by the second term on the right side where
M(ω)= 1/[1 + (τω)2]½
φ(ω) = - tan-1 τω
or these can be found from Figures 3.12 and 3.13. With ω = 2.5 rad/s,
M(2.5) = 0.37
φ(2.5) = -68o = -1.19 rad
Then,
ysteady(t) = (0.37)(1)(10)sin (2.5t + π/2 - 1.19)
= 3.7sin (2.5t + 0.38)
The time lag arising between input and output signals is given by
β1= φ(ω)/ω
= -1.19/2.5 = 0.48 s
COMMENT
The dynamic error in this problem is -63%, and this is a very large number. In effect, the
measurement system cannot respond quickly enough to follow the input signal. This creates a
filtering effect whereby a significant portion of the signal amplitude is attenuated (the term
"attenuation" refers to a reduction in value and is indicated by a negative dynamic error).
This system is a more effective as a filter than it is as a measuring system. Associated with
this large dynamic error is a large phase shift and associated time lag.
PROBLEM 3.11
KNOWN:
First order system
τ=2s
0.98 ≤ δ(ω) ≤ 1.02 required
FIND: The maximum frequency that can be measured ωmax
SOLUTION
The dynamic error is defined as
δ(ω) = M(ω) - 1
A first order system cannot have a value of M(ω) that is greater than 1.
So based on the constraint for dynamic error, we want
1 ≥ M(ω) ≥ 0.98
or
0.98 ≤ 1/[1 + (τω)2]½
At τ = 2 s, we find
ω ≤ 0.10 rad/s
For this to be true,
ωmax = 0.10 rad/s
or in terms of cyclical frequency
fmax = 2πωmax = 0.016 Hz
and
β1 = - 1.97 seconds
PROBLEM 3.12
KNOWN:
First order system
τ = 0.01 s = 10 ms
δ(ω) ≤ ±0.10
FIND:
M(ω), φ(ω). Frequency range to meet δ constraint.
ASSUMPTION: Input of the form, F(t) = A sin ωt
SOLUTION
For a first order system subjected to a periodic waveform input, the magnitude ratio and
phase shift are given by (3.10) and (3.9) respectively. For this particular system,
M(ω) = 1/[1 + (0.01τω)2]½
φ(ω) = - tan-1 0.01ω
or
ω [rad/s]
M(ω)
φ(ω)
1
10
100
1000
1.0
0.995
0.707
0.10
-0.6o
-5.7o
-45o
-84o
Note that at ω = τ, that M(ω) = 0.707, a value that defines the bandwidth of a system.
For a first order system, M(ω) will always be equal or less than unity. So that the dynamic
error constraint reduces to δ(ω) ≤ -0.1. For this to be true,
-0.1 ≥ (1/[1 + (0.01ω)2]½ - 1
solving,
ω ≤ 48.5 rad/s
It is clear that for all ω such that 0 ≤ ω ≤ 48.5 rad/s, the constraint is met.
PROBLEM 3.13
KNOWN:
First order system
τ = 0.15 s
K = 5 mV/oC
T(0) = 115 oC
F(t) = T(t) = 115 + 12sin 2t oC
ASSUMPTIONS: Output signal is linearly proportional to input signal (that is, K is
constant)
FIND: Output response, E(t) ; δ(ω); β1
SOLUTION
We see immediately from the units of static sensitivity that this first order instrument senses
temperature and outputs a voltage signal. Hence, a good system model can be written as
•
τ E + E = KF(t)
(3.4),
where E(t) represents the output voltage signal. Specifically, the system model is written as
•
0.15 E + E = 5(115 + 12sin2t) mV
with E(0) = KT(0) = 575 mV.
To solve for E(t), we assume a solution consisting of both homogeneous and particular parts,
E(t) = Eh + Ep
Because a first order system has only one root, the solution to the characteristic equation has
the form,
Eh = Ce-t/τ = Ce-t/0.15
We guess a particular solution of the form,
Ep = A + Bsin 2t + Dcos 2t
substitute back into the model with E(t) = Ep and
Ep = 2Bcos 2t - 2Dsin 2t
This leads to A = 575, B = 64.95 and D = 16.51. The full solution then has the form,
E(t) = Ce-t/0.15 + 575 + 64.95 sin 2t + 16.51 cos 2t
Evaluating at E(0) = 575 yields that C = -16.51. Lastly, the measurement system's output
response to this input can be rewritten as
E(t) = 575 + 67 sin(2t + 0.254) - 16.51e-t/0.15 mV
The dynamic error at ω = 2 rad/s can be expressed as
δ(2) = M(2) - 1
Using equation 3.10 to solve for M(ω) yields
M(2) = 1/[1 + ((2)(0.15))2]½ = 0.96
so that δ(2) = - 0.04.
The time lag can be found from
β1 (ω) = φ(ω)/ω
Using (3.9) at ω = 2 rad/s,
φ(2) = -tan-1 (2)(0.15) = -16.7o = 0.29 rad
so that the time lag is β1(2) = 0.29/2 = 0.15 s.
PROBLEM 3.14
KNOWN:
First order instrument
F(t) = 100U(t) oC
Five seconds are available to interpret signal and provide control signal back
to process.
FIND: τmax
SOLUTION
In such a situation, we will want the system to commence a shut-down when the output
signal achieves some set threshold value. We must set this threshold value. However, we
probably do not wish to set it too low or we run the risk of an unnecessary shut down
resulting from just random noise. Suppose we set the error fraction at 10%, that is at
Γ ≤ 0.10
reactor shut down will commence. Then with this value we can find the acceptable time
constant using
Γ = e-t/τ
With Γ = 0.10 and τ ≤ 5 s, we solve that τ ≤ 2.17 s.
COMMENT
We should note that as the set threshold value is pushed to lower values of error fraction,
the value for time constant becomes smaller. For example, at Γ = 0.01, τ ≤ 0.92 s. This places
a more restrictive design constraint on the sensor and installation selected.
PROBLEM 3.15
KNOWN:
Single loop LR circuit (see below) used as filter
R = 1million ohm
τ = L/R
FIND: L such that M(2000π) ≤ 0.5
SOLUTION
For the LR filter circuit shown, we can use the voltage loop law to write
L dI/dt + IR = Ei(t)
But across the resistor, I = Eo(t)/R. So the system governing equation is given by
L
•
E + E o = 1 E i (t)
R o
R
with Ei(t) = A sin ωt. This is a first order system which has a magnitude ratio defined as
M(2000π) = 1/[1 + (2000πτ)2]½ ≤ 0.5
Solving leads to τ ≤ 0.276 ms. From which
L = Rτ = (1 x 106Ω)(0.000276 s) = 276 H
L
Ei
R
Eo
PROBLEM 3.16
KNOWN:
F(t) = 2U(t)
ωn = 0.5 rad/s; ζ = 0.5; K = 0.5 m/V
•
y(0) = y(0) = 0
FIND: 90% rise time and settling time
SOLUTION
From (3.14a), for t ≥ 0+
y(t) = 0.5(2) − 0.5(2)e − 0.5(0.5)t [
where ωd = 0.43 rad/s.
By inspection,
90% rise time = 2.74 s
90% settling time = 9.5 s
0.5
1 − 0.5
2
sin0.43t + cos0.43t] = 1 − e − t/4 [0.58sin0.43t + cos0.43t]
PROBLEM 3.17
KNOWN:
Second order system
ζ = 0.6, 0.9, 2.0
FIND:
Plot M(ω) and φ(ω).
Find the frequency range in which δ(ω) ≤ ±5%
SOLUTION
The magnitude ratio and phase shift of a second order system is given by (3.21) and (3.19),
respectively,
M(ω)= 1/[(1 - {ω/ωn}2)2 + (2ζω/ωn)2]½
φ(ω) = tan-1 -(2ζω/ωn )/(1 - [ω/ωn ]2)
These are plotted below.
For a constraint of δ(ω) ≤ ±5%: 0.95 ≤ M(ω) ≤ 1.05.Either from the plots or from (3.21)
directly, for δ(ω) ≤ ±5%, then
ζ = 0.6; 0 ≤ ω/ωn ≤ 0.84
ζ = 0.9; 0 ≤ ω/ωn ≤ 0.28
ζ = 2.0; 0 ≤ ω/ωn ≤ 0.08
P3.17
10
0.6
0.9
Magnitude ratio
2
1
0.1
0.01
0.1
1
w/wn
10
PROBLEM 3.18
KNOWN:
Input: 30 ≤ C1 ≤ 40oC with f1 = 10 Hz
δ ≤ -0.01
FIND:
T(t), τmax
ASSUMPTIONS: First-order system (i.e. τ requested)
K = 1 unit/unit
SOLUTION
The input signal has the form:
T(t) = (40 + 30)/2 + [(40 - 30)/2] sin(2π(10Hz)t ±π/2)
= 35 + 5 sin(20πt ±π/2)
The dynamic error δ(ω) = M(ω) - 1 so setting M(ω) ≥ 0.99:
0.99 ≥ 1/[1 + (20πτ)2)½
τ ≤ 2.27 ms
PROBLEM 3.19
KNOWN:
Step test imposed on a second order system
Damped oscillating response, therefore ζ < 1
FIND: Test plan to find ωn and ζ
SOLUTION
The step test response for an underdamped system is plotted in Figure 3.14 (and the concept
further explored in the Example surrounding Figure 3.15). The period of oscillation gives Td
= 2π/ωd where
ωd = ωn (1 - ζ2)1/2
(1)
and the decay of the oscillations (i.e. the decay of the peaks of each cycle) follows e-ωn ζt. The
test plan should impose the step input and system output recorded at time intervals much less
than Td (e.g. 20 measurements per cycle). The peak values (maximum amplitudes), or
alternatively the minimum amplitudes, should be plotted versus time. Because the decay is
exponential, that is ymax = e-ωn ζt , a plot using semi-log will yield
ln ymax = ln (e-ωn ζt ) = -ωn ζt
where the slope of this plot (ln ymax vs. t) is
m = -ωn ζ
(2)
Equations 1 and 2 provide for the two unknowns.
3
2.5
y(t)
2
1.5
Td
1
0.5
0
0
5
time
10
PROBLEM 3.20
KNOWN:
Td = 0.577 ms from a step test
Second-order system due to oscillations observed
FIND:
ωd
SOLUTION
The period of oscillation of a step test is the period of ringing.
Accordingly, the ringing frequency is
ωd = 2π/Td = 1089 rad/s
or fd = ωd /2π = 173 Hz.
PROBLEM 3.21
KNOWN: ζ = 0.8
ωd = 2000π rad/s (where ω [rad/s] = 2πf [Hz] and f = 1000 Hz)
K = 1.5 V/V
F(t) = (12 + 24)/2 + [(24 - 12)/2]sin 600πt = 18 + 6sin 600πt
FIND: ysteady(t)
ASSUMPTIONS: Second order system
SOLUTION
The steady response to this input will have the form
ysteady(t) = KA + KAM(ω)sin ωt + φ(ω)
where
M(ω)= 1/[(1 - {ω/ωn}2)2 + (2ζω/ωn)2]½
φ(ω) = tan-1 -(2ζω/ωn )/(1 - [ω/ωn ]2)
The natural frequency is readily computed from the measured ringing
frequency by
ωd = ωn (1 - ζ2)1/2
yielding ωn = 3333π rad/s. Then at ω = 600π,
M(600π) = 0.99
φ(600π) = - 0.289 rad
Hence,
ysteady(t) = 27 + 8.9 sin 600πt - 0.289
COMMENT
Recall that the steady response is the output signal after all transients have died out. The
transient response is found from the homogeneous solution to the equation model.
PROBLEM 3.22
KNOWN: F(t) = A sin 500πt (recall ω [rad/s] = 2πf [Hz] and f = 250 Hz)
Constraint: δ ≤ ±0.02
Available ωn = 1200π rad/s (recall ω [rad/s] = 2πf [Hz] and f = 600 Hz)
Available values of 0.5 ≤ ζ ≤ 1.5
FIND: ζ required to meet the dynamic error constraint.
ASSUMPTIONS: Second order system
SOLUTION
For δ ≤ ±0.02 , we want 0.98 ≤ M(ω) ≤ 1.02. From the information given, the frequency ratio
is ω/ωn = 0.417. Then solving for ζ in the equation (3.19),
0.98 ≤ 1/[(1 - {ω/ωn}2)2 + (2ζω/ωn)2]½
yields
δ ≤ 0.72
Repeating at the other limit,
1.02 ≤ 1/[(1 - {ω/ωn}2)2 + (2ζω/ωn)2]½
yields
δ ≤ 0.63
Thus, either a transducer having δ = 0.65 or 0.7 will work.
PROBLEM 3.23
KNOWN:
Second order system responding to a sine wave as an input
2 ≤ f ≤ 40 Hz
M(f) ≤ 0.5
FIND: m, k, c
SOLUTION
Because we wish to attenuate frequency information at and above 2 Hz, we can initiate the
design of the system so that it meets the attenuation constraint at 2 Hz. Then we want,
M(f) = M(ω⁄2π) = 1/[(1 - {ω/ωn}2)2 + (2ζω/ωn)2 ]½ ≤ 0.5
A freebody diagram is constructed from a model below. The system model will have the
form,
••
•
m y + c y + ky = F(t)
so that, ωn = (k/m)1/2 ; ζ = c/2(mk)1/2; K = 1/k. Now at f = 2 Hz or ω = 2πf = 12.57 rad/s,
ω/ωn = 0.4. Use Figure 3.16 as a guide, note that for ω/ωn = 0.4 and M = 0.5, we will need
to have ζ < 2.5. Also from the Figure, it is apparent that this value will meet the constraint at
40 Hz as well. Suppose we select m = 2 kg as a reasonable estimate for the mass of a
turntable and select k = 4000 N/m, a value that allows the isolation pad to deflect 5 mm.
Then, the damping coefficient of the pad should be 447 N-s/m. Of course, the actual values
selected will depend on availability.
F(t)
F(t)
m(d2 y/dt2)
m
y
ky
Freebody diagram
c(dy/dt)
m
k
y
c
Idealized model
PROBLEM 3.24
KNOWN: RCL circuit: L = 2 H; C = 1 µF; R = 10kΩ
Ei (t) = 1 + 0.5 sin 2000t V
I(0) = dI(0)/dt = 0 (initial conditions)
ASSUMPTIONS: Values for R, C, L remain constant.
FIND: Governing equation and steady output form
SOLUTION
For a single loop RCL circuit, we apply the voltage law to the RCL loop,
E R (t) + E C (t) + E L (t) =
E i (t)
But,
ER= IR;
EL= L dI/dt; E C =
1
Idt
C∫
Substituting back into the loop equation and differentiating once gives,
dE
d2I
dI 1
+ R + I =i
dt 2
dt C
dt
This is of the form for a second order system. It follows that
L
=
ωn
1/=
LR ; ς R / 2 =
LC; K 1/ C
ωn = 707 rad/s; ς = 3.54; K = 1 x 10-6
From equations 3.21 and 3.19, respectively with ω = 2000 rad/s,
M(2000 rad/s) = 0.05
φ (2000 rad/s) = -1.23 rad
Then, with dEi / dt = 1000sin 2000t
[I(t)]steady = 1 + 0.025 sin(2000t - 1.23) µA
PROBLEM 3.25
KNOWN: Second order instrument
ς = 0.7
ωn = 2000 π rad/s (recall ω [rad/s] = 2πf [Hz] and f = 1000 Hz)
Input: 0 ≤ ω ≤ 1500π rad/s
FIND:
Does transducer meet a +/- 10% dynamic error constraint?
SOLUTION The dynamic error is defined by
δ(ω=
) M(ω) − 1
where for a second order system
M (ω ) =
1
1/ 2
2
 
2
2

 1 − (ω ωn )  + ( 2ζ ω ωn ) 




The dynamic error constraint can be rewritten as
0.9 ≤ M(0 ≤ ω ≤ 1500π) ≤ 1.10
Inspection of Figure 3.16 shows that the magnitude ratio will never exceed a value of 1.10
over this frequency range. However, its value can fall below 0.90. If we test the frequency
response at 1500 π rad/s, we can check this. Solving with for M( ω ) at ω / ωn = 0.75 and ς =
0.7 yields M(1500 π ) = 0.88.
This gives a dynamic error of -12%. So the transducer does not meet the constraint over
the entire frequency range of interest.
PROBLEM 3.26
KNOWN: t90 = 100 ms
fd = 1200 Hz
ς = 0.8
Recall ω [rad/s] = 2πf [Hz]; and so ω / ωn = f / f n , and so forth
FIND: δ (1 Hz), β1
SOLUTION
or f n 1200 / 1 − 0.82 = 2000 Hz
=
fd f n 1 − ς2 =
The dynamic error is given by δ(f ) = M(f) - 1:
M ( f= 1)=
1
1/ 2

2
2
 
 1 − ( f f n )  + ( 2ζ f f n ) 

 

2
=
1
1/ 2
2

2
2
 1 − (1 2000 )  + ( 2(0.8)1/ 2000 ) 


= 1.0
so δ (1 Hz) = 0.0.
The time lag is given by


Φ (ω ) Φ ( f ) 
−1 2ζ f f n 
β1 = = − tan
=
/ 2π f
2

ω
2π f
1 − ( f f n ) 

= 7.3 ms
COMMENT
The dynamic error of zero indicates that the amplitude of the input
signal KF(t) and the steady output signal are essentially equal. The time lag indicates that
the output signal appears at a time β1 after the input signal is applied.
PROBLEM 3.27
KNOWN: ωR = 1414 rad/s
ς = 0.5
f = 6000 Hz
Recall ω [rad/s] = 2πf [Hz]
FIND: δ (6000 Hz), Φ (6000 Hz)
SOLUTION
For resonance,
ωR =ωn 1 − 2ς 2
or 1414 =
ωn 1 − 2(0.5) 2 , so ωn = 2000 r/s
That means, fn = 318 Hz (because ω [rad/s] = 2πf [Hz])
The ratio f/fn = 6000/318 = 18.87 ; ς = 0.5 and so for f = 6000 Hz,
M(f ) =
1
1/ 2
2
 
2
2

 1 − ( f f n )  + ( 2ζ f f n ) 




or M(6000 Hz) = 0.0028.
The dynamic error,
δ (6000 Hz) = M(6000 Hz) - 1 = - 0.997
The phase shift is
Φ (6000 Hz) = − tan −1
2ζ f f n
1 − ( f fn )
2
2(0.5)(18.87)
=
=
− tan −1
2
1 − (18.87 )
3.04 rad = - 174o.
So the input signal amplitude is attenuated 99.7% at 6000 Hz with a 174o phase shift between
the input and output signals. The instrument effectively filters out the information at 6000
Hz.
PROBLEM 3.28
KNOWN: Second-order instrument
0 ≤ ω ≤ 100 rad/s
Constraint: δ(ω) ≤ ± 10%
FIND: Select appropriate values for ωn and δ
SOLUTION
The most demanding application will be at the highest input frequency
because δ(ω=
) M(ω) − 1 . Inspection of Figure 3.16 or use of equation 3.21 proves
useful to find:
ωn [rad / s] ς
200
200
200
500
500
500
0.4
1.0
2.0
0.4
1.0
2.0
ω
ωn
M(ω)
δ(ω)
0.5
0.5
0.5
0.2
0.2
0.2
1.18
0.64
0.47
1.03
0.96
0.80
0.18
-0.36
-0.53
0.03
-0.04
-0.20
So ς between 0.4 through 1 and with ωn = 500 rad/s will meet the constraint.
PROBLEM 3.29
KNOWN: First order system: τ = 0.2 s; K = 1 V/N
F(t) = sin t + 0.3 sin 20 t N (and so A1 = 1N and A2 = 0.3N)
FIND: Steady response output signal
SOLUTION
This system must respond to two frequencies ω1 = 1 rad/s and ω2 = 20 rad/s. The steady
output will have the form,
y=
KA1M(ω1 ) sin(ω1t + Φ1 ) + KA 2 M(ω2 ) sin(ω2 t + Φ 2 )
steady
For a first order system, we use equations 3.10 and 3.9 to solve for each M and Φ :
M(1 rad/s) = 0.98
Φ (1 rad/s) = -0.197 rad
M(20 rad/s) = 0.24
Φ (20 rad/s) = -1.326 rad
so that,
ysteady = 0.98 sin(t - 0.197) + 0.072 sin(20t - 1.326)
COMMENT
While the ω1 information is passed onto the output signal with a
minor attenuation (2%), the ω2 information is well attenuated (filtered) by 76%. This
measurement system is not a good choice for measuring the ω2 information.
PROBLEM 3.30
KNOWN: Second order system accelerometer:
ς = 0.5
(recall ω [rad/s] = 2πf [Hz])
f n = 4000 Hz
f = 2000 Hz
FIND:
δ(f ),f R
SOLUTION
For any system,
δ(f )= M(f ) − 1
where M ( f ) =
1
1/ 2

2
2
 
 1 − ( f f n )  + ( 2ζ f f n ) 

 

2
. Then, for this system
M(2000) = 1.11
so that
δ(2000) = + 0.11
The output magnitude from this system at this frequency will be 11% greater
than the input magnitude, KA. The system fails the criterion of +/- 10%.
The resonance frequency is found from
f=
f n 1 − 2ς 2 = 2828 Hz or about 2800 Hz.
R
PROBLEM 3.31
KNOWN: Second order transducer
ς = 0.4
f n = 18,000 Hz
(recall ω [rad/s] = 2πf [Hz] and f = 4500 Hz)
F(t) = A sin 9000 π t
FIND: δ(f ),f R , Φ (f )
SOLUTION
For any second order system,
δ(f )= M(f ) − 1
where M ( f ) =
1
1/ 2
2

2
2
 
 1 − ( f f n )  + ( 2ζ f f n ) 

 

. Then, for this system
M(4500 Hz) = 1.04
So output amplitude B is 4% higher than input amplitude A. This means the dynamic error
is
δ(4500Hz) = + 0.04
The phase shift is found from equation 3.19. For this system,
Φ (4500) =
− tan −1
2ζ f f n
1 − ( f fn )
2
=
− tan −1
2(0.4)(0.25)
1 − ( 0.25 )
2
= - 0.21 rad
That is about a 12o lag between input and output signals. The resonance frequency is found to
be
=
f R 18000 1 − 2(0.4) 2 = 14,843 Hz
PROBLEM 3.32
KNOWN: Seismic Accelerometer of Example 3.1
F(t) = x o sin ωt
FIND: M( ω ) and Φ (ω)
SOLUTION
From example 3.1,
 + cy + ky = cx + kx
my
which can be rewritten,
1
ωn2

y+
2ζ
2ζ
y + =
y
x + x
ωn
ωn
= y(0)
Assuming zero value initial conditions, y(0)
= 0
(ω
y (t=
) yh +
ωn ) sin [ωt + Φ (ω ) ]
1/ 2
2

2
2
 
 1 − (ω ωn )  + ( 2ζ ω ωn ) 




Then,
(ω
M (ω ) =
ωn )
1/ 2
2

2
2
 
 1 − (ω ωn )  + ( 2ζ ω ωn ) 




Φ (ω ) =
− tan −1
2ζ ω ωn
1 − ( ω ωn )
2
By inspection, this instrument will be best suited to measure signals having frequency
content that is greater than its natural frequency.
COMMENT
The instructor may wish to augment this problem with material from Section 12.5 or refer
students to this section of the text for further reading.
PROBLEM 3.33
KNOWN: Second order system pressure transducer
ς = 0.6
ωn = 8706 rad/s
FIND: M(ω), Φ(ω), ωR
SOLUTION
For a second order system the magnitude ratio and phase shift are
M (ω ) =
1
1/ 2
2

2
2
 
 1 − (ω ωn )  + ( 2ζ ω ωn ) 




Φ (ω ) =
− tan −1
2ζ ω ωn
1 − ( ω ωn )
2
Using the known values, we construct the frequency response as
1.0
1.0
1.04
0.83
0.01
Φ(ω) [rad]
0
-0.14
-0.72
-1.57
-3.02
For underdamped systems, the
maximum value for M(ω) occurs
at ωR,
1.2
0
-0.5
1
-1
0.8
-1.5
0.6
-2
0.4
-2.5
0.2
-3
0
ωR =ωn
1 − 2ς 2
= 4607 rad/s
Phi(w)
10
1000
4607
8706
87000
M(ω)
M(w)
ω [rad/s]
1
10
100
1000
10000
-3.5
100000
w
so M (4607) = 1.042. The resonance is slight for ζ = 0.6 as verified in the magnitude ratio
plot in the text.
PROBLEM 3.34
KNOWN: Coupled first and second order systems
F(t) = 10 + 50 sin 628t oC
Input Stage Device:
τ = 1.4 ms
K = 2 V/oC
Output Stage Device
K = 1 V/V
ς = 0.9
ωn = 10000 π rad/s
FIND: Steady portion of output signal y(t)
SOLUTION
For a coupled system of two components, call them 1 and 2, the system output is defined
by
=
KG ( s ) K=
G ( s ) K 2 G2 ( s ) K1 K 2 M1M 2 e
1 1
with
i ( Φ1 +Φ 2 )
M (ω ) system = M1 (ω ) M 2 (ω )
Ksystem = K1K2
Φ system = Φ1 (ω ) + Φ 2 (ω )
For ω = 628 rad/s,
M1 (628 rad/s) =
M 2 (628 rad/s) =
1
{1 + (ωτ ) }
2 1/ 2
= 0.75,
1
1/ 2
2

2
2
 
 1 − (ω ωn )  + ( 2ζ ω ωn ) 




= 1.0,
Φ1 (628 rad/s) =
− tan −1 ωτ = -0.721 rad,
2ζ ω ωn
Φ 2 (628 rad/s) =
− tan −1
2
1 − ( ω ωn )
= -0.036 rad
Then, with Ksystem = (2V/oC)(1 V/V) ,
ysteady(t) = (10)(2)(1) + (50)(2)(1)(1)(.75)sin(628t + (-0.721-0.036)
= 20 + 75 sin(628t - 0.757) V
PROBLEM 3.35
KNOWN: Two coupled second order systems
Input signal: 2 ≤ C1 ≤ 5mm and f1 = 85 Hz
Constraint: δ(ω) ≤ ± 0.05
FIND:
ωn , ζ for each of the two measurement system stages
SOLUTION
There are a number of ways to approach this problem and there is not one answer. The
following is one approach to choosing a system.
The input function has the form,
F(t) = (5+2)/2 + [(5-2)/2]sin 170πt mm
= 3.5 + 1.5 sin 170πt mm
Now in order to set the specifications, we need to examine how the system as a
whole will respond to the input of frequency 170π rad/s. For coupled systems,
=
KG ( s ) K=
G ( s ) K 2 G2 ( s ) K1 K 2 M1M 2 e
1 1
with
M (ω ) system = M1 (ω ) M 2 (ω )
Ksystem = K1K2
Φ system = Φ1 (ω ) + Φ 2 (ω )
In general, for second order systems,
M 2 (ω ) =
1
1/ 2
 
2
2

 1 − (ω ωn )  + ( 2ζ ω ωn ) 

 

Φ 2 (ω ) =
− tan −1
2
2ζ ω ωn
1 − ( ω ωn )
2
The dynamic error of the system is found from
δ(ω)system= M(ω)system − 1
i ( Φ1 +Φ 2 )
If we accept, δ(ω) ≤ ± 0.05 as a maximum constraint, then this means
0.95 ≤ M1M 2 ≤ 1.05
So choose any value for M1 and M2 that is within this constraint.
For example, we could take,
M1 = 0.98 and M2 = 1.02
as one possible design target. With these values selected, we examine each stage in the
system.
Input Stage Device
Suppose we fix ζ1 = 0.7, then for M1 (170π) = 0.98, ωn = 81π rad/s. The
1
phase shift then becomes, Φ1 (170π) = -0.71 rad.
Output Stage Device
Suppose we fix ζ 2 = 0.6, then for M 2 (170π) = 1.02, ωn = 48π rad/s. The
2
phase shift then becomes, Φ 2 (170π) = -0.35 rad.
COMMENT
The actual values for ζ and ωn may be limited due to availability from
vendors. But the problem demonstrates one approach to dealing with such an
open ended problem. Note also that the phase shift for the system selected is about
-0.61 rad, which is well tolerated for most purposes and is within the range in
which phase is linear with frequency.
PROBLEM 3.36
KNOWN: ωR = 82.5 rad/s
ζ= 0.4
K = 2 V/N
F(t): Ao=3 N, C1=C2=1 N, ω1 =8 rad/s = 1.27 Hz, ω2 = 165 rad/s = 26.26 Hz
FIND: y(t)
SOLUTION
y(t) = yh + 3K + KC1M(8) sin[8t + φ(8)]
+ KC2M(165) sin[165t + φ(165)]
For this system the natural frequency is
ωn = ωR /(1 - 2ζ2 )1/2 = 82.5 r/s /(1 - 2(0.4)2)1/2 = 100 rad/s
With ω1 /ωn = 0.08 and ω2 /ωn = 1.65 and ζ = 0.4, use of Figure 3.16 and 3.17 or equations
3.21 and 3.19 give:
M(8 rad/s) = 1.004 (using equations)
M(165 rad/s) = 0.46
φ(8 rad/s) = -3.7o
φ(165 rad/s) = -142.5o
The transient response yh is given by equation 3.14a and appropriate initial
conditions.
y(t) = yh + 6 + 2 sin[8t - 3.7o] + 0.92 sin[165t - 142.5o]
Program Sampling (companion disk) was used to generate the amplitude spectrum for y(t)
Problem 3.36 (cont)
PROBLEM 3.37
KNOWN: First stage: τ1 = 0.10 s, K1= 1 V/V
Second stage: K2 = 100 V/V, fn= 15000 Hz, ζ = 0.8
Input signal: F(t) = 5 sin 2000 t [mV]
FIND: y(t), ζ (f ), β1
SOLUTION
y(t) = yh(t)+ K1K 2 M1M 2 *5sin(1000t + Φ1 + Φ 2 ) mV
For ω = 1000 rad/s:
M=
(1000rad/s)
1
1
=
2 1/ 2
1 + (ωτ )
{
}
1
{1 + (1000*0.1) }
2 1/ 2
= 0.995
-5.7o
Φ 2 (ω ) =
− tan −1 ωτ − tan −1 (1000*0.100) =
For ω / ωn = 0.42 and Φ = 0.8:
M 2 (1000) =
1
1/ 2
2
 
2
2

 1 − (ω ωn )  + ( 2ζ ω ωn ) 




Φ 2 (1000) =
− tan −1
2ζ ω ωn
1 − ( ω ωn )
2
= 0.95
= -39.2o
so:
y(t) = yh(t) + 0.467 sin[1000t - 44.9o] mV
COMMENT
This output is amplified by the second stage (K2 = 100V/V) .
A second stage with a higher natural frequency would bring M2 closer to unity and decrease
the phase shift.
PROBLEM 3.38
KNOWN: frequency bandwidth 0 to 20,000 Hz ±1dβ
FIND: Translate this specification into words
SOLUTION
A typical audio amplifier increases the output amplitude relative to the input
amplitude by some amount and that amount is called its gain. You may be more familiar with
the term ‘power’ instead of gain, such as in the expression 100 Watts power. This power is
simply another way to state the gain.
Now in our equations, the gain is just the static sensitivity, K, for the amplifier at
some reference frequency. So the amplifier gain is stated at some reference frequency. In
literature pertaining to amplifiers, the static sensitivity is called the static gain – but it all
means the same thing.
This system specification states that for an input frequency within 0 to
20,000 Hz the amplifier gain does not vary by more than 1dβ . Another way to write this is:
−1dβ ≤ KM(0 ≤ f ≤ 20000Hz) ≤ 1dβ where KM is the output amplitude. So the product
KM(f) is frequency dependent and therefore the amplifier gain is frequency dependent.
Now, from the definition of decibel, + 1dβ is equivalent to M(f) = 1.12 or δ(f ) =
+0.12 or a 12% increase over the reference amplifier gain. The - 1dβ is equivalent to M(f) =
0.89 or δ(f ) = -0.11 or a 11% decrease over the reference amplifier gain. So between 0 and
20,000 Hz, the signal amplitude is essentially constant to within ±1dβ . A typical audio
amplifier will have some spikes and troughs across its frequency response. But the
specification is explicit that the amplitude never varies by more than the 1dB.
Music signals are a series of sinusoidal frequency terms. Even single
notes, such as a middle C, consist of a fundamental frequency and harmonics.
The harmonics give distinction to the source of the sound so that different
instruments are recognizable. Under normal circumstances, we would want the
reproduction electronics to neither add nor detract from the signal
information (i.e. we want M(f) = 1 across the spectrum).
PROBLEM 3.39
KNOWN: Input Stage Device
K1 = 10 mV/mm
ωn = 10000 rad/s
1
ζ1 = 0.6
Output Stage Device
K2 = 1 mm/mV
ωn = 700 rad/s
2
ζ 2 = 0.7
ysteady(t) = 90 sin(4πt + Φ1 ) + 50sin(80 π t + Φ 2 )
FIND: Determine if measurement system specifications are adequate for
the input signal.
SOLUTION
Both devices are second order systems:
M (ω ) =
1
Φ (ω ) =
− tan −1
1/ 2
2
 
2
2

 1 − (ω ωn )  + ( 2ζ ω ωn ) 

 

2ζ ω ωn
1 − ( ω ωn )
2
For the output signal given, the input signal must have the form
F(t) = 90/KM(4) sin 4 π t + 50/KM(80 π ) sin 80 π t
For coupled systems,
=
KG ( s ) K=
G ( s ) K 2 G2 ( s ) K1 K 2 M1M 2 e
1 1
Then for second order systems,
M1(4 π ) = 1
Φ1 (4π) = 0
M1(80 π ) = 1
Φ1 (80π) = 0
M2(4 π ) = 1
Φ 2 (4π) = -0.02 rad
M2(80 π ) = 0.99
Φ 2 (80π) = - 0.52 rad
Hence,
Msystem(4 π ) = 1
Msystem(80 π ) = 0.99
i ( Φ1 +Φ 2 )
Φ system (4π) = -0.02 rad
Φ system (80π) = -0.52 rad
Ksystem = 10 mm/mm
The excellent response characteristics of the measurement system make it a suitable choice
for this measured signal.
PROBLEM 3.40
KNOWN: F(t) = A1 sin170πt + A 2 sin 254πt + A3 sin 904πt
Input Stage Device Availability
1000π ≤ ωn ≤ 2000π rad/s
ζ = 0.5
Output Stage Device
δ(0.1 ≤ f ≤ 250kHz) ≤ -3 dB
FIND: Select an acceptable value for ωn
SOLUTION
There are numerous approaches and solutions to this problem. We offer one possible
solution to illustrate the design selection approach.
The second order displacement transducer will be most heavily tested at the highest input
frequency. Suppose we impose the restriction on the transducer that δ1 (ω) ≤ ±0.1 where we
let δ1 (ω) be the dynamic error due to the transducer. Then, for a second order device,
ωn
M(904 π ) M(254 π ) M(170 π )
[rad/s]
1000 π
1500 π
1750 π
2000 π
ωR
[rad/s]
1.03
1.14
1.12
1.09
1.03
1.01
1.01
1.01
1.01
1.00
1.00
1.00
707 π
1060 π
1237 π
1414 π
But lets check this performance out. A quick inspection of Figure 3.16 shows that the
input transducer will experience some resonance behavior if ωn is too close to the input
frequencies. For example, with ωn = 1000 π rad/s, the 904 π rad/s input frequency drives the
transducer into the post peak resonance region of the graph. That is not good.
So it would better to select a transducer where ω << ωR . To do this, raise the natural
frequency. So we select ωn = 2000 π rad/s. Now we should meet our criterion without
resonance problems. The spectrum measurement device will not be a factor owing to its wide
frequency response relative to these input frequencies.
PROBLEM 3.41
KNOWN:
Three Amplitudes at three times
ωd = 10 rad/sec
FIND:
ωn and ζ
SOLUTION
Three amplitudes of A1 = 17 mV, A16 = 9 mV, and A32 = 5 mV occur at times:
t1 = Td/2 = 0.3142 s, t16 = 15Td+t1 = 9.739 sec, and t32 = 31Td + t1 = 19.792 s
The amplitudes decay at the rate given by:
ymax = e-ωnζt
Taking the log of both sides:
ln (ymax) = ln(e-ωnζt) = -ωnζt
If plotted, the slope of this relation will be: m = -ωnζ
The data follow the relation: Y = -0.06X + 2.838. The slope is –0.06.
ωnζ = 0.06
ωn = ωd/(1 - ζ2) = 10/(1 - ζ2)
Solving simultaneously gives: ωn = 10.01 rad/sec; ζ = 0.006
P3.41
ln (amplitude)
3
2.5
2
1.5
1
0
10
20
time [s]
30
PROBLEM 3.42
This problem exercises the ability to articulate an understanding of the concepts of static
sensitivity, natural frequency and damping ratio relative to a measurement system. The short
essay should be written in the student’s own words rather than a restating of text material.
Each student understands material in their own individual way. So this is an opportunity for
written technical communication between instructor and student.
PROBLEM 3.43
KNOWN: RL circuit (only one reactive element here, therefore first order)
R=4Ω
L = 0.1H
Ei (t) = 50 V for t > 0
t(0-) = 0
FIND:
I(t)
SOLUTION
At t = 0, the switch closes applying the 50V potential across the RL circuit. This is a
step function input as the circuit sees it, i.e., Ei(t) = 50U(t) V where Ei(t) = 0 for t < 0- and
Ei(t) = 50 V for t > 0+ . Around the loop:
Ei (t) − RL − L
dI
=
0 t>0
dt
or
L
1
t>0
I + I = Ei (t)
R
R
This is of the form τy + y =KF(t) , so by
inspection: τ =L / R ; K = 1/R, F(t) = Ei(t).
I(t)=
Ei (t)
+ (0 −
Ei (t)
)e− tR / L = 12.5 − 12.5e − t / 0.025 for t > 0+
R
R
= 12.5(1 − e − t / 0.025 ) amps
for t > 0+
COMMENT
In practice, a halfwave rectifier D1 with a free-wheeling diode D2 would be added to the
circuit to avoid an inductive voltage spike that can arise when applying a current surge to an
inductor.
PROBLEM 3.44
KNOWN: RC circuit (it is first order because there is only one reactive element)
R = 1 k ohm
C = 1000 µ F
EB = 6 V ( = Ei(t) for t > 0+)
Ec(0) = 0 (this assumes capacitor is initially totally discharged)
FIND: t90 time for capacitor to reach 90% of its maximum energy level.
SOLUTION
For a flash circuit initially at zero potential, the closing of the switch is a step function
change in voltage, i.e. Ei(t) = EBU(t) and Ei(t < 0-) = 0 and Ei(t > 0+) = EB.
Lets start by estimating the total energy that can be stored in the capacitor,
1
1
=
CE c2
CE 2B = (0.5)(1000x10-6 F)(6 V)2 = 18 x 10-3J
2
2
so, 90% of this amount is e90 = (0.9)(18 x 10-3J) = 16.2 x 10-3J. For the capacitor to reach this
1
energy level, its voltage Ec is e90 = CE c2 or Ec = 5.692 V. So we seek the time required to
2
charge the capacitor to this voltage level.
=
e
For the RC circuit for t > 0, we want to examine the capacitor voltage as a function of time as
driven by the applied battery voltage:
1
1
E c −
Ec −
E =
0
RC
RC B
or
RCE c − E c =
EB
But this is of first order form
τE − E =
KF(t)
c
c
So, τ =RC = (1000 Ω)(1000 x 10-6 F) = 1 s, and K = 1 V/V, and F(t) = EB.
Γ(t) = 0.1 =
E c (t) − 6V
= e− t / τ = e− t
0 − 6V
with Ec(t) = 5.692 V, then t90 = 2.97 s.
PROBLEM 3.45
Program Thermal Response plots the time response from a first order device. The input
magnitude is controlled by the user and may be varied with time. The instrument time
constant is set by the user and may be varied with time. The user should select a time
constant and an amplitude, start the program, and then vary the amplitude – creating a step
change in input.
The system response slows down (relative to time) as the time constant is increased
independent of the magnitude of the step change imposed.
PROBLEM 4.1
KNOWN: N > 1000; x = 9.2 units ; Sx = 1.1 units
FIND: Range of x in which 50% of all measurements should fall.
ASSUMPTIONS: Measurand follows a normal density function
Data set sufficiently large such that x ≈ x’ and Sx ≈ σ
SOLUTION
By assuming that the data is sufficiently large such that its population behaves as an infinite
population, we can find the interval defined by
x' - z1 σ ≤ xi ≤ x' + z1 σ
as follows. We can find P(x' + z1 σ ) from the one-sided integral solution to
P(z1 ) =
1
[2π]
1/2
Z1
∫e
−β 2 /2
dβ
0
This solution is given in Table 4.3 for p(z1) = 0.25 (one half of the 50% probability sought)
as z1 = 0.674. Then, we should expect that 50% of the xi values lie in the interval given by
9.2 – 0.7425 ≤ xi ≤ 9.2 + 0.7425 (50%)
COMMENT
We can see from Table 4.4 that as N becomes large the value for t approaches a value given
by z1.
4-1
PROBLEM 4.2
KNOWN: N > 10 000; x = 204 units ; Sx= 18 units
FIND: x' - z1σ ≤ x ≤ x' + z1σ at P = 90%
ASSUMPTIONS Measurand follows a normal density function
Data set sufficiently large such that x ≈ x’ and Sx ≈ σ
SOLUTION
Using the definition z1 = (x1 - x')/σ we find the z1 value corresponding to the one-sided
probability integral p(z1) = 0.45 from Table 4.3. This gives,
z1 = 1.65
Then,
1.65 = (x1 - x')/ σ
or
1.65σ = x1- x'
or
x1= x' + 1.65σ
But a normal (Gaussian) distribution is symmetric about the mean value. Hence, for 90%
probability (2)(0.45),
x' - 1.65σ ≤ x ≤ x' + 1.65σ
(90%)
or
174.3 ≤ x ≤ 233.7 units (90%)
4-2
PROBLEM 4.3
KNOWN:
x = 121.6 psi
Sx = 14 psi
N is very large
FIND: P(x > 150 psi)
ASSUMPTIONS:
Normal distribution
Sx ≈ σ ; x ≈ x'
SOLUTION
The z variable is defined by
z1 = -(x1 - x')/ σ = -(121.6 -150)/14 = 2.028
We look up P(2.028) from Table 4.3. Interpolation gives
P(2.028) = 0.4786
This expresses the probability that 121.6 ≤ x ≤ 150 psi. Then, the probability
that x > 150 psi is
0.5 - 0.4786 = 0.0214
or there is a 2.14% probability that any measurement will yield a value in excess of 150 psi.
4-3
PROBLEM 4.4
KNOWN: Toss of four coins
FIND: Develop the histogram for the outcome of any toss.
State the probability of obtaining three heads on one toss.
SOLUTION
A coin has two distinct sides, so each toss has two possible outcomes. With four coins,
there will be 24 = 16 possible outcomes of any one toss of the four coins. The probability of
three heads is 4 in 16 or 25%. The possible outcomes are:
Number
of Heads
nj
4
3
2
1
0
1
4
6
4
1
The histogram is shown below. Because of the few number of tosses, the development of the
histogram is primitive. But the symmetry is obvious. This type of distribution is best
Possible Occurrences, n
Problem 4.4
8
6
4
2
0
0
1
2
3
4
Number of Heads in a Toss
described as a Binomial distribution (see Table 4.2).
COMMENT
The binomial distribution shape is similar to the Gaussian (normal) distribution, except that it
can lack the extended "tails" found with the Gaussian shape. As the number of possible
outcomes (number of coins tossed) becomes large (say 30 or more), the two distributions
become nearly identical over a wide interval about the mean and the Gaussian distribution
can be used for ease.
4-4
PROBLEM 4.5
SOLUTION
In the matchbox game, the frequency distribution will more closely resemble a Gaussian
(normal) distribution as N becomes larger. This is because each shot is independent of the
other and each shot differs from the other by random variation.
A 'better' player will have a mean distance in the outcome that is close to the target point and
have a low variance. That is, the player will have a low systematic error (average distance
from target point) and low random error (variations from the average point).
This game and its interpretation are similar to the dart game discussed in Chapter 1 in the
discussion of random and systematic error. In the US, a variation of this game is called
'matchbook football' where the objective is to slide the matchbook across a table so that it
just overhangs the table edge.
4-5
PROBLEM 4.6
FIND: Histogram for Table 4.8, Column 1.
SOLUTION
A distribution is not unique and we give one possible solution.
For N = 20 values, K = 7 is selected. The interval (bin) values are
Bin #
1
2
3
4
5
6
7
Interval
range
<49
49 - 49.6
49.6 - 50.2
50.2 - 50.8
50.8 - 51.4
51.4 - 52
>52
Number of Occurrences
Column 1
6
5
4
3
2
1
0
<49
49.6
50.2
50.8
Force
4-6
51.4
52
>52
PROBLEM 4.7
FIND: Frequency distribution for Table 4.8, Column 3.
SOLUTION
A distribution is not unique and we give one possible solution.
For N = 20 values, K = 7 is selected. The interval (bin) values are
Bin #
1
2
3
4
5
6
7
Interval
range
<49
49 - 49.6
49.6 - 50.2
50.2 - 50.8
50.8 - 51.4
51.4 - 52
>52
Frequency of Occurrence
Column 3
0.3
0.25
0.2
0.15
0.1
0.05
0
<49
49.6
50.2
50.8
51.4
Force
4-7
52
>52
PROBLEM 4.8
FIND: Compare and discuss histograms for Table 4.8, Column 1,2 and 3.
SOLUTION
A distribution is not unique and we give one possible solution.
For N = 20 values, K = 7 is selected. The interval (bin) values are
Bin #
1
2
3
4
5
6
7
Interval
range
<49
49 - 49.6
49.6 - 50.2
50.2 - 50.8
50.8 - 51.4
51.4 - 52
>52
Number of
Occurrences
Colum n 2
6
5
4
3
2
1
0
<49
49.6 50.2
50.8 51.4
Force
52
The variations seen are likely a consequence of (1) normal variation due to finite data sets, (2) random errors in
each measurement. Each histogram clearly shows a central tendency and in each case it is in bin 4.
4-8
>52
PROBLEM 4.9
KNOWN:
Three datasets from Table 4.8 Column 1,2, and 3
N = 20
FIND: F , SF, and S F
SOLUTION
The mean value for each dataset is
F=
F1 = 50.5 N
1 N
∑F
N i=1 i
F2 = 50.7 N
F3 = 50.6 N
The standard deviation for each dataset is
=
SF [
1 N
( Fi − F ) 2 ]1/ 2
∑
N − 1 i =1
with ν = N - 1 = 19 for each individual dataset.
S F = 1.00 N
1
S F = 1.21 N
2
S F = 1.01 N
3
The standard deviation of the means expected based on each individual dataset is
S =
SF
( N )1/ 2
S F = 0.22 N S F = 0.27 N
F
1
1
S F = 0.23 N
1
The standard deviation of the means value simply reflects the possible random error in the
mean value of the dataset relative to the true mean value of the data population due to data
scatter.
4-9
PROBLEM 4.10
KNOWN:
Data from Table 4.8 , Column 1,2, and 3
N = 20
SOLUTION
Dataset:
1
2
3
Minimum: 48.9 48.7 48.9 N
Maximum: 52.4 52.6 52.4 N
Means:
50.5 50.7 50.6 N
While each dataset shows a range (maximum – minimum) of possible values, the data clearly
show a preference for a value in the range 50.2 to 50.8 N (or bin 4). This is what is meant by
a central tendency in a population – a tendency for a data point to have or be close to one
value over all others.
Number of
Occurrences
Colum n 1
6
5
4
3
2
1
0
<49
49.6 50.2 50.8 51.4
Force
52
>52
52
>52
Number of
Occurrences
Colum n 3
6
5
4
3
2
1
0
<49
49.6 50.2 50.8 51.4
Force
4-10
PROBLEM 4.11
KNOWN:
Data of Table 4.8, Column 1
N = 20
F ± tSF (95%)
FIND:
SOLUTION
For this dataset in Column 1:
1 N
=
∑ F 50.5 N
N i=1 i
=
F
=
SF [
1 N
1/ 2
1.00 N
∑ ( F − F )2 ]=
N − 1 i =1 i
with ν = N − 1 = 19 . Then from Table 4.4, t19,95 = 2.093. The scatter of the measured data set
can be expressed by
Fi= F ± t19,95 S F (95%) = 50.5 ± (2.093)(1.00) = 50.5 ± 2.1 N (95%)
1
Here Fi denotes the value of any measurement of force, F.
Column 2:
Fi= F ± t19,95 S F (95%) = 50.7 ± 2.53 N (95%)
2
Column 3:
Fi= F ± t19,95 S F (95%) = 50.6 ± 2.11 N (95%)
3
COMMENT
The above probability statement reflects the scatter of the data set. In effect, it provides a
range of values in which any measured value is expected to fall with 95% probability. A 95%
probability with this statistical statement implies that at least 19 out of every 20
measurements are expected to fall in the range defined for Fi. Indeed, inspection of the
dataset shows that this is a true statement.
4-11
PROBLEM 4.12
KNOWN:
Data of Table 4.8, Column 1
N = 20
FIND:
F ± tS
(95%)
F
SOLUTION
For this dataset in Column 1:
=
F
1 N
=
∑ F 50.5 N
N i=1 i
=
SF [
=
S
F
1 N
1/ 2
( Fi − F ) 2 ]=
1.00 N
∑
N − 1 i =1
SF
=
1.0=
/ 20 0.22 N
N1/2
with ν = N − 1 = 19 . Then from Table 4.4, t19,95 = 2.093. Then, we can expect that the true
mean value should lie within the interval defined by
F ± tS
(95%)=50.5 ± 0.47 N (95%)
F
This gives the mean value for this data set and a statement of the range of mean values we
would expect to find from any data set. A 95% probability indicates that at least 19 out of
every 20 complete data sets should show a sample mean value within the range. Compare the
meaning of this statement to that found in Problem 4.11. They are very different!
Column 2:
F ± tS
(95%)=50.7 ± 0.57 N (95%)
Column 3:
F ± tSF
(95%)=50.6 ± 0.47 N (95%)
F
In all three datasets, the mean values fall within an overlapping ring, as predicted.
COMMENT
The reasoning behind this confidence interval for the mean value lies within the limitations
of finite statistics. While the sample mean value defines the mean value of the 20 data points
exactly, it is not necessarily the true mean value of the measured variable (compare the
4-12
results for these three real data sets in columns 1,2, and 3 – each has a different sample
mean). Different data sets of the same variable will give somewhat different mean values. As
N becomes large, the sample mean will approach the true mean and the confidence interval
will go to zero. Remember this assumes that there is no systematic error acting on the
measurement.
4-13
PROBLEM 4.13
KNOWN: Data of Table 4.8, Column 3
One additional measurement
FIND:
F − t 95SF ≤ F21 ≤ F + t 95SF
SOLUTION
We can show that
Fi= F ± t19,95 S F (95%) = 50.6 ± 2.11 N (95%)
3
where Fi denotes the value of any measurement of force, F. Hence, an additional
measurement of F would be expected to fall within the range of 48.49 to 52.71 with 95%
probability (likelihood).
4-14
PROBLEM 4.14
KNOWN: Data of Table 4.8, Columns 1, 2 and 3.
N = 19 (repetitions)
M = 3 (replications)
F ± t 95 S
FIND:
F
ASSUMPTIONS: Data sets in Columns 1,2 and 3 represent duplicate data sets of the
same measured variable under similar operating conditions.
SOLUTION
We can find the pooled mean value of the combined data sets
< F >=
1
MN
M
N
∑∑
j=1 i =1
Fij =
1
M
3
∑ F j = 50.6N
j=1
Where M = 3 and N = 20. The pooled standard deviation
< S F >= (
1
M(N − 1)
M
N
∑ ∑ (Fij − Fj ) 2 )1/2 = 1.06N
j=1 i =1
The pooled standard deviation of the means is
< S >=
F
< SF >
(MN)1 / 2
= 0.14 N
with degrees of freedom, ν = M(N-1) = 57. From Table 4.4, t57,95 = 2.00. Then, the best
estimate of the true value is given by
F' =< F > ± < tS >= 50.6 ± 0.28 N
F
(95%)
COMMENT
We can see that the confidence interval for the ensemble mean value is reduced over any one
dataset. While column 3 predicts the ensemble mean value the closest, all three are well
within the confidence band. We learn that the more information (measurements) we have
about a dataset, the more confident we become in their true values. Yet, very reasonable
estimates can be made from smaller datasets. The number of measurements, required
confidence, and the cost of additional measurements all must be carefully weighted.
4-15
PROBLEM 4.15
KNOWN: Table 4.8, Column 1 dataset
x = 50.5N
Sx = 1 N
FIND: Test to determine if the dataset follows a normal (Gaussian) distribution
SOLUTION
With x ' ≈ x and σ x ≈ Sx , we test. The data lends itself to 7 intervals:
j
1
2
3
4
5
6
7
interval
0 - 49
49 - 49.6
49.6 – 50.2
50.2 – 50.8
50.8 – 51.4
51.4 – 52
52 – above
nj
1
3
4
5
4
2
1
nj’
1.34
2.35
3.96
4.71
3.96
2.34
1.34
Σ=
(n j − n 'j ) 2 / n 'j
0.0863
0.1823
0.0004
0.0179
0.0004
0.0494
0.0863
0.423
As an example, consider j = 2, where the number of observations is n2 = 3. The expected
number of outcomes for a normal distribution is estimated to be 2.346, as follows:
P(49 < x i < 50.5) − P(49.6 < x < 50.5)
= P(z a ) − P(z b )
x − x ' 49 − 50.5
x − x ' 49.6 − 50.5
=
= −1.5
zb =
=
= −0.9
σ
1
σ
1
so, P(z a ) − P(z b ) = 0.4332 − 0.3159 = 0.1173
za =
for N = 20, nj’ = 20 * .1173 = 2.346. So we expect ~ 2.35 occurrences and we observe 3:
Then, (n 2 − n '2 ) 2 / n '2 = 0.182. This is the deviation between expected and observed.
For 7 intervals and using calculated values of the 2 statistical quantities, x and Sx, we have
ν = 7 − 2 = 5 degrees of freedom.
For χα2 (5) =
0.423 , α ~ 0.995. So, P( χ 2 ) = 0.005 < 0.05 So the test is unequivocal: the
distribution is a normal distribution.
4-16
PROBLEM 4.16
KNOWN: x’ = 100 N
σ2 =400 N2 (so, σ = 20 N)
FIND: For N = 16, P(90 ≤ x ≤ 110) =
?
SOLUTION
Begin by finding the z value for a corresponding x
x −x'
z=
σ/ N
90 − 100
For x = 90 N, z =
= -2.0
20 / 16
For x = 110 N, z =
110 − 100
= 2.0
20 / 16
So, P(90 ≤ x ≤ 110) = P(−2.0 ≤ z ≤ 2.0) = 0.9544
So, there is about a 95% chance.
4-17
PROBLEM 4.17
KNOWN: Large sample of grades (i.e. infinite statistics applies)
FIND: Number of A, C, D grades awarded (per 100 students)
SOLUTION
Referring to the probability graph below:
A: P(1.6) = 0.4452 so that the area under the 'A' region is:
0.5 - P(1.6) = 0.0548. Hence, 5.48% (or between 5 and 6) are A's.
C: P(0.4) + P(0.4) = 2P(0.4) = 0.3108
Hence, 31.08% (or about 31) are C's
D: P(1.6) - P(0.4) = 0.4452 - 0.1554 = 0.2898
Hence, 28.98% (or about 30) are D's.
Likewise, 5.48% are F's and 28.98% are B's.
4-18
PROBLEM 4.18
KNOWN:
x' = 20 µm; σ = 30 µm
FIND:
P(xi ≥ 80µm ); P(50 ≤ xi ≤ 80 µm )
SOLUTION
P(xi ≥ 80µm) = P(z ≥ z1)
So computing z1: z1 = (80 – 20)/30 = 2
Then, P(z ≥ 2) = 0.0228 or a 2.28% chance (1 in 44).
P(50 ≤ xi ≤ 80 µm ) = P(za ≤ xi ≤ zb)
So computing: za = (50 – 20)/30 = 1.0
zb = (80 – 20)/30 = 2.0
P(za) = 0.1587
P(zb) = 0.0228
So: P(50 ≤ xi ≤ 80 µm ) = P(za ≤ xi ≤ zb) = 0.1587 – 0.0228 = 0.1359 or about a 14% chance
(that’s about 1 in 7).
4-19
PROBLEM 4.19
KNOWN:
N = 10; xi
FIND:
x ; Sx2 ; P(x ≤ 203 µm)
ASSUMPTION: Data represents infinite population
SOLUTION
Based on the data given, we compute:
x = 209.6 µm
Sx2 = 2753.1 µm2
Sx = 52.5 µm
Using these values as representative of an infinite population:
P(xi ≤ 203µm) = P(z ≤ z1)
z1 = (203 – 209.6)/52.5 = 0.126
P(z1) = 0.05
Or about 5% of the data fall between the mean value and 203µm. Since 50% of the data lie
above the mean value, then
P(z ≤ z1) = 0.5 – P(z1) = 0.45 or about 45% will fall below 203 µm.
4-20
PROBLEM 4.20
KNOWN: p(x) where x is in hours
FIND: mean value of x
SOLUTION
The mean value can be found from the known probability
density function:
∞
x
=
∞
xp(x)dx ∫ 0.001xe−0.001x dx
∫=
−∞
−∞
= 1000 hrs.
The average expected life of the bulb is 1000 hours (60,000 minutes).
4-21
PROBLEM 4.21
SOLUTION
In the absence of systematic error, the sample mean value estimates the true mean value
with a confidence interval given by ± tS =
± t(Sx / N ) . The value of S represents the
x
x
standard random error in the mean value.
As the number of measurements, N, increases, the confidence improves, i.e. the interval gets
smaller at a rate of 1/N1/2. There is some gain as the value of t drops as N becomes larger, but
N is essentially 2 after 30 measurements and approaches an asymptotic value of 1.96 as N
goes to infinity.
Hence, the indicator S in N = 16 measurements improves to only twice that of N = 4
x
measurements despite quadrupling the number of measurements:
Sx / 16 versus Sx / 4
Likewise, increasing N from 25 to 100 only halves the value of S .
x
COMMENT
For small sample sizes, the gain in this random error is impressive for making
some extra measurements. But as N increases, this gain requires considerably more
measurements. Doubling N from 10 to 20 is more efficient than increasing N from 1000 to
2000. This is the "diminishing returns" in using N to improve random error.
4-22
PROBLEM 4.22
KNOWN: N = 270 with x = 6.92 MN/m2 and Sx2 = 6.89 (MN/m2)2
FIND: tS at 95%
x
SOLUTION
The true mean value of these bricks is given by
x' = x ± tS (95%)
x
With N = 270, t 269,95 = 1.96, so with Sx= [6.89 (MN/m2)2]1/2 = 2.62 MN/m2
± tS =
±(1.96)(2.62MN / m 2 ) / 270 =
0.313MN / m 2
x
This is the estimate for the random error in the mean value due to variation in the dataset.
We can state that
=
x ' 6.92 ± 0.313MN / m 2 (95%)
Based on the dataset (and in the absence of other errors), there is a 95% probability that x'
lies between 6.61 and 7.23 MN/m2.
4-23
PROBLEM 4.23
KNOWN:
N = 61
x = 44.20 N
Sx2 = 4 N2
FIND: P(45.56 ≤ x ≤ 48.20 N)
SOLUTION
The t value is defined by t ≡ (x − x) / Sx where Sx = 2 N. With this
ta = (45.56 - 44.20)/2 = 0.68
We use Table 4.4 but must recognize that it is a two sided t chart (includes equal area on both
sides of the mean value) and so we must correct to a one-sided value. For ν = N - 1 = 60, we
find P ≈ 50% or 0.50. So that P(44.20 ≤ x ≤ 45.56 N) = 0.5/2 = 0.25. Similarly,
tb = (48.20 - 44.20)/2 = 2.0
Again, from the two sided t chart (Table 4.4) at ν = N - 1 = 60, P ≈ 95% or 0.95. So
that P(44.2 ≤ x ≤ 48.2 N) = 0.95/2 = 0.475. Then,
P(45.56 ≤ x ≤ 48.20 N) = 0.475 - 0.25 = 0.225
So there is a 22.5% chance that a measured value of x will fall within this
interval.
4-24
PROBLEM 4.24
x = 924.2 MPa
Sx = 18 MPa
KNOWN:
N = 36
x ± tS
FIND:
x
SOLUTION
=
S 18
=
/ 36 3.0
x
For ν = 36 -1 = 35 degrees of freedom, . So in the absence of other errors, the
estimate of the mean shear strength is
x ± tS = 924.2 ± 6MPa (95%)
x
The 95% confidence interval is 918.2 to 930.2 MPa for mean shear strength.
4-25
PROBLEM 4.25
KNOWN: M = 3 pooled data sets
FIND: ν, x , and x ± tS
x
(95%)
ASSUMPTIONS: The three data sets are replicate measurements of a variable under similar
conditions.
SOLUTION
For pooled statistics of a single variable with M = 3 replications,
M
∑ ν j=
ν=
M
∑ (N j − 1)=
15 + 20 + 8= 43
=j 1 =j 1
The weighted pooled mean value is
x = (32 + 30 + 34) / 3 = 32units
The pooled standard deviation is
1/ 2
S=
[(15*32 + 20* 22 + 8*62 ) /(15 + 20 + 8)]=
3.42units
x
The standard deviation of the means is
S = 4 / (16 + 21 + 9)= 0.59 ≈ 0.6units
x
Then, t43,95 ~ 2.018 (interpolation) or just use 2.0 (as a general approximation for N >30).
x'=
32 ± (2.018)(0.59) =
32 ± 1.2units (95%)
COMMENT
In this problem we are faced with three somewhat different results
obtained from measuring the same variable on three separate occasions. The
variations in the statistics between each data set reflect (1) the ability to
duplicate the operating conditions for each test exactly, and (2) the limitations
of finite statistics. Please review "replication" discussed in Chapter 1.
4-26
PROBLEM 4.26
KNOWN:
x = 3027 psi
N = 11
Sx = 53 psi
FIND: Is x' ≥ 3000 psi at 95% probability
SOLUTION
In the absence of other errors, we know that
x =' x ± tS
x
(95%)
For N = 11, ν = 10 and t10,95 = 2.228 . The standard deviation of the means for this data set
is
S = 53 / 11 = 15.98 psi
x
Then,
=
x ' 3027 ± 35.6psi (95%)
or there is a 95% probability that the true mean strength of the footing is in the interval
2991 ≤ x ' ≤ 3063
The data suggest the possibility that the footing does not meet the code with a
95% probability. In particular, we know that the footings are prepared in
sections as the concrete trucks arrive, unload, and depart. Therefore, portions
of the footing likely do not meet the code for strength. Oops - Time to break up and
repour!
4-27
PROBLEM 4.27
KNOWN:
Data set of N = 10 values
FIND: Check for outliers.
ASSUMPTIONS: Fixed operating conditions.
Measured variable has a normal distribution.
SOLUTION
The statistics for this data set are found to be
x
=
N
xi
∑=
923.7 N
i =1
1.2
 1 N

Sx 
(x i − x)  = 8.13 N
=
∑
 N − 1 i =1

The modified three sigma test introduces the modified z variable, zo
=
z o (x i − x) / Sx
A visual survey of the data indicates that data point #3 could be suspect. Computing
a value for zo with xi = 908 N gives, zo = 1.93. From Table 4.3, P(1.93) =
0.4732. Then,
N[0.5 - P(1.93)] = 0.27
Since this value exceeds 0.1, the value for data point #3 apparently falls within
the bounds to be expected from normal scatter and is NOT an outlier. No
outliers are detected in this data set.
Using t9,95 = 2.262, and Sx = Sx / N = 2.57,
x' = 923.7 ± 5.8 N (95%)
4-28
PROBLEM 4.28
KNOWN: N = 20 measurements taken from a large batch
Sample statistics: x = 47.5 mm Sx = 8.4 mm
Claim: x' = 42.1 mm (for batch)
FIND: Is claim supported by the data set?
ASSUMPTION: Sample is representative of the batch.
SOLUTION
For N = 20, ν = 19 so that at the 95% level, t19,95 = 2.093. The true
mean based on the batch statistics is
x' = 47.5 ± (2.093)(8.4)/201/2= 47.5 ± 3.93 mm
The claim is not supported by the sample.
4-29
(95%)
PROBLEM 4.29
KNOWN:
x' = 37.84 mm ; σ = 0.13 mm
FIND:
P(37.58 ≤ x ≤ 38.1 mm)
SOLUTION
P(37.58 ≤ x ≤ 38.1 mm) = P(za ≤ x ≤ zb)
za = (38.1 – 37.84)/.13 = 2
zb = (37.58 – 37.84)/0.13 = -2
P(za) = 0.4772
P(zb) = 0.4772
Recognizing that za and zb are on opposite sides of the mean value,
P(za ≤ x ≤ zb) = P(za) + P(zb) = 0.9544 or about 95%.
4-30
PROBLEM 4.30
KNOWN:
Data set provided.
N=5
FIND: Best curve fit to the data set and 95% confidence interval.
SOLUTION
The best curve fit is the one that best fits the physics of the problem. Not knowing
that, we try several curve fits below.
The data can be fit to: yc = ao + a1x + ... + amxm
m ao
a1 a2
a3
t ν ,95 Syx tSyx
ν
1 1.90 1.37 ---- ---3
3.182 0.648 2.06
2 2.70 0.44 0.17 ---- 2
4.303 0.37 1.59
3 1.89 2.25 -0.71 0.11 1 12.706 0.09 1.14
The best fit would depend on the problem physics. But a third order fit reduces tSyx to a
minimum and appears to fit the data well. The data are plotted below.
12
This first plot shows the curve fit
and its uncertainty band based on
±tS yx . For most engineering
2
0
0
∑ ( xi − x )2 = 25.29
i =1
and x = 6.16. The differences
between the two results are nearly
indistinguishable (hence, we
plotted them separately).
1
2
3
4
5
6
X
12
10
y = 0.11x 3 - 0.71x 2 + 2.25x + 1.89
8
Y
N
6
4
This second plot shows the curve
fit and its uncertainty band based


2 

(x − x )
1
 . Here
on ±tS yx  + N i
N

∑ ( xi − x )2 

i =1


the value of
y = 0.11x 3 - 0.71x 2 + 2.25x + 1.89
8
Y
problems where the error in y is
much greater than the error in the
controlled variable x, this
approach gives reasonable
numbers and is quick and easy.
10
6
4
2
0
0
1
2
3
X
4-31
4
5
6
PROBLEM 4.31
KNOWN:
Data set provided.
N=6
FIND: Best curve fit to the data set and 95% confidence interval.
SOLUTION
Using a least squares regression, a third order polynomial fits the dataset
y = 1.8245x3 - 10.12x2 + 44.097x
better than the second order polynomial y = 17.219x2 - 54.826x + 46.779.
(Incidentally, when using a spreadsheet, such as ExcelR , the polynomial can be estimated by
plotting the data, clicking on the data on the plot, and setting the trendline option to the
polynomial order desired - or you can attempt another transformation.)
1400
1200
Y
1000
y = 17.219x 2 - 54.826x + 46.779
800
600
400
200
0
0
2
4
6
8
10
X
Looking at this data set, it is attractive to try a transformation of the form
log y = m logx + log b or Y = mX + B
where Y = log y; X = log x and B = log b. This is equivalent to yc = bxm.
The data are plotted below on a log-log plot and fit to the curve yc = 5.05x2.3. For t4,95 =
2.770 and Syx = 0.025, the confidence interval is shown on the log-log plot.
4
log y
3
2
1
0
-0.5
0
0.5
log x
4-32
1
1.5
PROBLEM 4.32
KNOWN:
Data set provided.
N=4
FIND: Best curve fit to the data set and 95% confidence interval.
SOLUTION
The data are plotted below and fit to the curve yc = 0.586x2.04.
The t value is found to be t2,95 = 4.303 and Syx = 0.021 and shown on the log-log plot.
100
80
60
y
y = 0.586x 2.04
40
20
0
0
2
4
x
6
8
10
2
log y
1.5
1
0.5
0
-0.5
-1
-0.5
0
0.5
log x
1
4-33
1.5
PROBLEM 4.33
KNOWN: Fan test data for Q and h
FIND: h = f(Q)
SOLUTION
Linear regression is used to find h = f(Q) as an mth order polynomial of the form:
h = ao + a1Q + a2Q2 + ... + am Qm
To keep the regression coefficients reasonable, the polynomial will be based on Q in
thousands of cms, i.e. kcms. A first order fit does not appear reasonable, so it is not
attempted.
m
ao
a1
a2
a3
r
2 4.9886 0.2906 -0.0211 ------- 0.9987
3 5.2991 0.1468 -0.0064 -0.0004 0.9999
Syx
t95Syx
0.1246 0.396
0.0329 0.013
While an inspection of tSy shows that the third order polynomial fits this data best, the
problem physics require that we use the second order polynomial – that is the best fit based
on known fan curves. Note how the r value is not very sensitive here and is generally not a
good indicator of how 'well' data fits a curve.
7
h (cm water)
6
5
4
2
h = -0.0211Q + 0.2906Q + 4.9886
3
2
1
0
0
5
10
15
Q (kcms)
4-34
20
25
PROBLEM 4.34
KNOWN: Constraint: want σ 2 ≤ 1.25 x10−4
S x2 = 2.1x10−4 based on N = 30 for a large batch
FIND: Is rejection of the batch based on S x2 prudent?
SOLUTION
With ν = 29
χα2 (ν) =νS2x / σ2 =(29)(2.1x10-4)/(1.25x10-4) = 48.7
2
Inspection of Table 4.5 shows, χ0.015
(29) ≈ 48.7 , so α ≈ 0.015 .
So there is about a 1.5% chance that this batch actually meets the
constraint despite the S2x value of the sample. Rejecting this batch is
prudent (or at least prudent with 98.5% confidence!).
4-35
PROBLEM 4.35
KNOWN: x = 5.060 mm with Sx = 0.0025 mm based on N = 30 measurements
Constraint: want x' = 5.000 mm based on N = 10,000 units.
FIND: Does the sample mean support the intention so that the constraint is met?
SOLUTION
In bearing grinding operations, the tooling setup determines the mean diameter of the
finished product.
For ν = 29, t29,95 ~ 2.045. The confidence interval of the mean based on the variation in
the dataset (and no other errors) is given by ± tSx / N =
± tSx ~ 0.001mm. This data set
suggests
5.059 ≤ x ' ≤ 5.061 mm
(95%)
No, the constraint is not being met for bearing diameter. Stop the machine and reset the
tooling set-up.
COMMENT
Clearly there is a systematic error in the bearing grinding setup causing the bearing diameter
mean value to be greater than the intended bearing diameter. The problem is not in the
random error (random variations in diameter between bearings). The whole setup is shifted to
making too large a bearing.
4-36
PROBLEM 4.36
KNOWN: Constraint: CI = ± 0.010 mm
S1 = 0.0025 mm based on N1 = 30
FIND: NT
SOLUTION
When the confidence interval ± tSx / N =
± tSx based on a reasonable number of
measurements (such as 30 or more such that the t value does not change
appreciably with increased N) is met, no further measurements are necessary.
Here we want ± tSx ≤ 0.01 mm and, in fact, ± tSx ≈ 0.001 mm . So the criterion is met.
So NT = N1.
4-37
PROBLEM 4.37
KNOWN:
Failure time for N = 6 pumps
FIND:
Mean and its 95% confidence interval
Number of specimens tested to reduce confidence in mean value to within 50
hrs.
ASSUMPTIONS: σ is approximated by Sx
SOLUTION
i.) Based on the data set provided:
N
x = ∑ x i = 1372.2 hrs
i =1
N
∑ (x
S x = ( i =1
i
− x) 2
N −1
) 1/2 = 125.6 hrs
S x = S x /N 1/2 = 51.3 hrs
Here t5,95 = 2.571.
So the true mean value is estimated by x =' x ± tSx :
x' = 1372.2 ± 131.8 hrs. (95%)
ii.) To reduce the interval of 131.8 hrs to an interval of 50 hrs at 95% confidence,
NT = (
t N −1,95S x
d
) 2 = 167
So a first guess is that 167 additional measurements would need to be taken. This would be
tested again following these measurements.
4-38
PROBLEM 4.38
FIND: x,Sx Test the hypothesis of a normal distribution.
SOLUTION
Tons
421-480
481-540
541-600
601-660
xj
nj
n'j
450.5
510.5
570.5
630.5
4
8
12
6
2.9
9.6
7.8
4.2
N
∑ xi
=1
x i=
=
N
(nj - n'j)2/ n'j
0.42
0.27
2.26
0.77
4
∑ n jx j
j=1
= 550.5 MPa
N
1/ 2
 4

 ∑ n j x 2j − Nx 
j=1

Sx = 


N −1




= 57.53 MPa
Do the data support the hypothesis of a normal distribution?
The expected occurrences n'j are listed above. For example, for the first interval:
za = (421 - 550.5)/57.53 = 2.25
zb = (480 - 550.5)/57.53 = 1.23
P(421 ≤ x i ≤ 480) = P(2.25) - P(1.23) = 0.4878 - 0.390 = 0.098
So for N = 30:
n'1 = (30)(.098) = 2.9 whereas we observe n1 = 4.
For all the intervals:
=
χα2
K
∑ (n j − n 'j )2 / n 'j
= 3.72
j=1
Then, for K = 4 here: ν = 2, and χα2 (2) =
3.72. Interpolation of Table 4.5 or using
a more extensive math handbook table, we find that α ≈ 0.85. There is a 85% chance that the
discrepancy between nj and n'j is due to random variation alone. Or we can look at this as P =
1 - α = 0.15, a 15% chance it is not due to random variation. This result is equivocal, that is
the hypothesis is possible and is not disproved: the sample could follow a normal
distribution.
4-39
PROBLEM 4.39
KNOWN: For N1 = 30, x = 550.5 MPa, S1 = 57.53 MPa
FIND: NT required to attain CI = ± 0.03 x
SOLUTION
CI = ± (0.03)(550.5 MPa) = ± 16.52 MPa
For d = CI/2 = 16.52 MPa,
(
)
2
N T = t N −1S1 / d = [(2.042)(57.53)/16.52]2 = 51
An additional 21 measurements should be taken to meet the constraint. The statistics should
then be recomputed to verify that the constraint is met.
PROBLEM 4.40
KNOWN: For N1 = 6, x = 71,327 psi and S1 = 8345 psi
FIND: NT required to achieve CI within 0.05 x (total range)
SOLUTION
For a total range of (0.05)(71,327 psi) = 3566 psi, CI = ± 1783 psi. With
d = CI/2 = 1783 psi:
NT = {(6)(8345)/(1783)]2 = 145
An additional 139 measurements are required to reach this constraint level of random error.
Because N1 is quite small relative to NT, it would be prudent to reevaluate the sample
statistics after some intermediate number of measurements were taken.
4-40
PROBLEM 4.41
KNOWN:
CI = 0.1 g
Sx = 2 g
FIND: N
SOLUTION
Let d = CI/2 = 0.05 g. We are looking for the number of measurements required
to keep tSx ≤ 0.05g at 95%.
N = ( tSx / d)2
If we select a large number of measurements, such that tN,95 = 1.96, then
N ≈ 6150
For this value, the t value remains unchanged. Thus, a large number of measurements are
required due to the close restriction on CI.
4-41
PROBLEM 4.42
KNOWN: N1 = 60
Sx1 = 1.52 V
CI = 0.28
FIND: NT
ASSUMPTIONS: 95% confidence required.
Sx1 is representative of σ .
SOLUTION
With N1 = 60, ν = 59 and t59,95 = 2.00. Setting d = CI/2 = 0.14, then
(
)
2
N T = tSx / d = [(2)(1.52)/0.14]2 = 472
As a first estimate, at least 472 total measurements will be needed to achieve
the desired precision levels. Hence, another 412 measurements are needed. This CI is a tight
constraint on random error given the large variations seen in the dataset (Sx1 value).
PROBLEM 4.43
FIND: P(χ 2 (10) > 19.0
SOLUTION
From Table 4.5: 0.05 ≤ χ 2 (10) ≤ 0.025
that is, the probability lies between 2.5 and 5%.
4-42
PROBLEM 4.44
KNOWN: Expect 0.07 breaks per meter.
FIND: p(x) for a wire of length L = 5 m.
ASSUMPTION: Model using poisson distribution
SOLUTION
From the given information, the most probable number of breaks to be
expected over the 5 m length is:
λ = x' = (0.07)L = (0.07)(5) = 0.35
p(x) = 0.035x e−0.35 / x!
x
p(x)
1
2
3
4
10
0.247
0.043
0.005
0.0004
0.0000
For example, there is a 4.3% probability of finding 2 breaks over the stated length, L.
4-43
PROBLEM 4.45
KNOWN:
2 out of every 100 screws are defective
x is the number of defectives.
FIND: p(x)
SOLUTION
The binomial distribution models the number of occurrences x of a defective part in N =
100 observations assuming the probability of finding either a defective part or a nondefective
part with any observation remains the same (that is, removing a part does not alter the
remaining population of defective parts). For N = 100, P(x) = 0.02. So pb will take the form:

 x


N!
100!
x
100 − x
=
p b (x) 
P (1 − P) N − x = 

 0.02 (1 − 0.02)
−
−
(N
x)!x!
(100
x)!x!




The poisson distribution models the events occurring randomly over a number of
observations. pp (x) refers to the probability of observing x defects in just 100 observations.
Take λ = x' = 2 as the number of expected defects in each 100 parts over an infinite
population:
p(x) = 2 x e−2 / x!
(Hint: use a spreadsheet or programmable calculator to solve each – those factorials get
large): For x defective parts in any 100 samples
x
pb
pp
0
1
2
4
10
0.1326
0.2707
0.2734
0.0902
0.0000
0.1353
0.2707
0.2707
0.0902
0.0000
Or there is about a 27% chance you will find 2 defective parts, a 13% chance of finding no
defective parts, but it is unlikely that you will find 10 defective parts.
COMMENT
Under certain circumstances the poisson distribution will approximate
the binomial distribution. This is because the former is the limiting case of the
latter, a case often proven in statistics texts. When N → ∞ and P → 0 such that
the product of N*P is constant, the two are exact. We see that the results above come close to
this condition.
4-44
PROBLEM 4.46
KNOWN:
Particle passage through a small volume
Average particles passing in time period t1 is 4: λ = 4
FIND: p(x) using a poisson distribution model
SOLUTION
Using λ = x' = 4 as the most probable expected value,
p(x) = 4 x e−4 / x!
x
p(x)
1
2
3
4
5
10
0.0733
0.1465
0.1954
0.1954
0.1563
0.0053
For example, there is a 15.63% chance of observing 5 particles in the defined time period, t1.
4-45
3.5
3
gy
2.5
2
PROBLEM 4.47
Finite Population estimates the histogram and finite statistics of an internally generated signal
representing a random variable. The signal is continually measured and the statistics and
histogram are constantly updated.
The histogram tends towards a Gaussian distribution as N becomes larger.
The values of the finite statistics change as new values are added to the data set. However,
clearly there is a tendency towards a mean value of zero. The maximum and minimum values
stay between +3 and -3 most (actually 95%) of the time.
PROBLEM 4.48
Running Histogram estimates the histogram of an internally generated signal. The user can
vary the number of intervals K and the number of data points N used in each histogram. The
program updates the histogram for each block of N data points.
As interval number is increased, the number of observations (magnitude) of each interval
decreases. But as the number of samples N is increased the signal increases (with time), the
histogram evolves towards a Gaussian shape. With the appropriate number of intervals for
the N samples, a very accurate portrayal of the sample tendency is found.
Values “out of range” are the random and rare occurrences that have values outside of the
limits of the histogram. In most cases, these could be considered as “outliers.”
PROBLEM 4.49
Probability Density estimates the probability density function of an internally generated
signal. The user can vary the number of intervals K and the number of data points N used in
each plot. The program creates a new pdf each time the program is rerun.
The signal has its own statistics and pdf, but each sample is but a snapshot of finite length of
the infinite signal. The pdf will change with each new data set because the finite statistics of
the data set (finite population) do not exactly estimate the statistics of the infinite data set
(infinite population).
4-46
“Learn by writing” methods are valuable for developing understanding of a new concept.
Questions 5.1 – 5.4 provide some topics for “Learn to Write” exercises.
PROBLEM 5.1
SOLUTION
Systematic error is a constant error that shifts all measured values of a variable by a fixed
amount. In effect, the sample mean value will be offset from the true mean value by this
fixed amount. Randomization methods break up some of the trends brought on by
interference, a result of systematic errors, Randomization makes systematic errors behave as
random errors, which are more easily quantified using statistics. Calibration is an excellent
way to isolate, identify, quantify and thereby reduce systematic errors.
Random error leads to scatter in the measured values obtained during the measurement of a
variable under otherwise fixed operating conditions. Unlike systematic errors, random errors
will change in magnitude between repeated measurements bringing on the noted scatter.
Both systematic and random errors are present in any measurement to some degree. For the
engineer, the difficult task is assigning the probable values of these errors. That’s where the
test plan comes into importance. By incorporating repetition into a test plan, random errors
can be statistically estimated with some amount of predictability. Incorporating replication
strengthens the random error estimates and allows estimates of control to be made, which
may include some of the systematic errors.
PROBLEM 5.2
SOLUTION
Systematic errors are usually estimated by comparison methods. These methods include: (i)
calibration, (ii) concomitant methods, (iii) interlaboratory or different facility comparisons,
or (iv) experience.
Random errors are manifested by measured data scatter and their effects on the estimate of
the true value of the measured variable can be estimated statistically using the methods
discussed previously in Chapter 4.
ERRORS ARE NOUNS; UNCERTAINTIES ARE NUMBERS. An error refers to a
difference between the measured value and the true value. Because this exact amount is
unknown, a probable range of the error is estimated. This numerical estimate is the
uncertainty.
PROBLEM 5.3
SOLUTION
TRUE VALUE: The actual value of the measured variable. The value
sought by measurement. Most often refers to the true mean value of the
variable which would result from an infinite sampling under perfect test
control.
BEST ESTIMATE: The nearest approximation of the true value that
can be made with the data set available. It is based on the data set and the
precision and the bias errors involved in the measurement. It is usually offered
by the sample mean value and qualified by its precision interval.
MEAN VALUE: Exact statement of the mean or central tendency of a
measured data set. The mean value of a finite data set is given by its sample
mean value.
UNCERTAINTY: Estimate of the precision and bias errors involved in
estimating the value of a variable. It is the range of probable errors which
affect the outcome of a measurement.
CONFIDENCE INTERVAL: The range (or interval) of values within
which the true value is expected to lie with some probability. The confidence
interval is in part based on the precision-based interval ( tSx ) discussed in Chapter 4 due
solely on the variations in the measured data set, but it will also include all other random
errors and systematic errors involved in the measurement.
PROBLEM 5.4
KNOWN: Tire pressure gauge; single sample uncertainty estimates
FIND: Difference between ud and uN
SOLUTION
The design-stage uncertainty is a first estimate that is based on information immediately
available. Generally, it does not include measurement control estimates, which makes it very
different from an Nth order analysis. The simplest form for ud is an estimate by
=
ud
u o2 + u c2
where uo is based on the expected instrument interpolation error and uc is based on the
instrument error. It can also include other sources of error known at the time of analysis. This
value provides a good guess of the uncertainty to be expected when minimal information is
available.
Nth order uncertainty is applied to those measurement situations where a number of repeated
measurements cannot or will not be taken. The analysis focuses on the control of the test and
how that would affect the outcome. The Nth order uncertainty includes all conceivable
errors, but its most important distinguishing feature is that it estimates the effect of test
process control on the measurement. It is estimated by
=
uN
N −1
u c2 + ∑ u i2
i =1
where ui are uncertainties related to the measurement procedure controllability.
Suppose we want to measure the pressure of a tire and make only one reading per tire during
a normal tire installation or during a tire pressure check. This is the way we tend to do it –
isn't it? How good might we expect that one value of pressure to be? To quantify our
methodology, we conceive of a test where we try out our measurement technique with a
surrogate set of measurements. So u1 might be estimated by a simple surrogate experiment
whereby a fixed pressure is repeatedly measured (say 20 times) and the outcome stated as
tSp . Assuming that the pressure did not change, variations in measured pressure would be
due to measurement procedure control, as well as instrument repeatability. This gives us an
estimate for how the pressure might be affected during our normal single measurement. We
bank this information for subsequent use. The values for ud and uN differ by the errors which
enter during the conduct and control of the test.
PROBLEM 5.5
KNOWN:
Micrometer
Resolution: 0.001 inch (0.025 mm)
FIND: ud at 95% confidence
ASSUMPTIONS: Instrument error is negligible compared with error due to
resolution.
SOLUTION
The interpolation error due to instrument resolution of an analog instrument is
approximated by half its least increment:
uo = 0.0005 inch or 0.0125 mm
It is not unusual for this type of instrument to have mostly negligible instrument errors.
However, its zero set point can only be controlled to within the precision of the resolution, so
we set uc ~ uo.
Then, the design-stage uncertainty becomes
ud =
±(uo2 + uc2 )1/ 2 = ± 0.0007 inch (95%) = ± 0.0180 mm (95%)
PROBLEM 5.6
KNOWN:
Analog Tachometer
Resolution: 5 rpm
Accuracy: within 1% reading
FIND: ud at 10, 500, 5000 rpm
SOLUTION
The design-stage uncertainty is
ud =
±(uo2 + uc2 )1/ 2
where
uo = ± 2.5 rpm
uc = 1% of reading
This yields
speed
[rpm]
uc
uo
ud
[rpm] [rpm] [rpm]
10
0.1
2.5
± 2.5
500
5
2.5
± 5.6
5000
50
2.5
± 50
The uncertainty increases with rotational speed. At low speeds it is dominated by the ability
to read the tachometer (resolution). At higher speeds the instrument errors dominate.
COMMENT
The statement "accuracy" is a manufacturer catch-all term that is rarely clearly defined.
The "accuracy" statement is presumed to mean that the overall errors do not exceed 1%. We
suspect the term to describe the combined effects of all known elemental errors. This
misnomer causes confusion. Insist on a detailed description.
PROBLEM 5.7
KNOWN: Speedometer
Resolution: 5 mph (8kph)
Accuracy: within ± 4% reading
FIND: ud at 60 mph (90 kph)
SOLUTION
The design-stage uncertainty is
ud =
±(uo2 + uc2 )1/ 2
where
uo = ± 2.5 mph (4 kph)
uc = ± 4% of reading = ± 2.4 mph at 60 mph = ± 3.6 kph at 90 kph
This yields
ud =
±(2.52 + 2.42 )1/ 2 = ± 3.5 mph (95%) = ± 5.4 kph (95%)
COMMENT
In the United States, automobile speedometers and their connected odometers have
acceptability tolerance limits set by the government at u = ± 4% of the reading.
During the tight fuel availability times in the western hemisphere of the 1970's, some
automakers were known to misrepresent (legally) automobile performance by holding a
tighter precision (lower random error) on their units while purposely building a 2 to 4%
systematic error into their automobile odometers. This made the vehicle slightly more
attractive to a fuel cost conscious consumer. Unless the consumer actually made a careful test
of the speedometer and odometer performance, usually by some comparison means such as a
road sign check, they never would have reason to disbelieve their car's indicated, but
inaccurate, performance. Caveat emperor!
PROBLEM 5.8
KNOWN:
Temperature sensor
Error limit: ± 0.5oC
Readout Device
Resolution: 0.1oC
Accuracy: 0.6oC
FIND: ud
SOLUTION
The design-stage uncertainty is for the combined system is
ud =
±[(ud ) 2R + (ud ) 2S ]1/ 2
where (ud)R is the design-stage uncertainty of the readout device and (ud)s is
that of the sensor. In either case, the individual design-stage uncertainty is found from
ud =
±(uo2 + uc2 )1/ 2
Sensor
uo = 0 (i.e. no output stage in the sensor; readout is separate)
uc = ± 0.5oC
( ud =
)S
(02 + 0.52 )1/ 2 = 0.5oC
Readout Device
uo = 0.05oC
and
uc = 0.6oC
So that
ud )
(0.052 + 0.82 )1/ 2 = 0.8 oC
(=
R
Then, the design-stage uncertainty for this combined system becomes
ud =
±(0.52 + 0.82 )1/ 2 =
± 0.8 oC (95%)
COMMENT
Although the error limit on the sensor was given in this problem, similar information is
available through various publications – an example of estimating error by using published
values. For example, there are readily available ASTM and ASME standards (e.g.,
ASME/ANSI PTC 19.3; also given in the textbook as Table 8.5) governing the error limits
on thermocouples, which are common temperature sensors, and to which a manufacturer's
product must adhere.
PROBLEM 5.9
Four resistors are available: two rated at R = 500 ± 50 Ω and two
rated at R = 2000 ± 5% Ω .
KNOWN:
FIND: Best design combination to form RT = 1000 Ω
SOLUTION
We can combine the resistors in series or in parallel. Consider as Case 1, a series
arrangement and as Case 2, a parallel arrangement.
Case 1
RT = R1 + R2
If we use the two 500 Ω resistors, then
ud = ± 50 Ω
R1
ud = ± 50 Ω
R2
The propagation of uncertainty through to RT is estimated by
∂R
∂R
(ud ) R =
±[( T ud ) 2 + ( T ud ) 2 ]1/ 2 = ±[(1* ud ) 2 + (1* ud ) 2 = ±71Ω (95%)
T
R1
R2
∂R1 R1
∂R2 R 2
Case 2
RT =
R1 R2
R1 + R2
If we use the two 2000 Ω resistors, then
ud = ± 100 Ω
R1
ud = ± 100 Ω
R2
The propagation of uncertainty through to RT is estimated by
∂R
∂R
(ud ) R =
±[( T ud ) 2 + ( T ud ) 2 ]1/ 2
T
∂R1 R1
∂R2 R 2
= ±[({
R2
R1 + R2
−
= ±35Ω (95%)
R1 R2
( R1 + R2 )
2
}* ud ) 2 + ({
R1
R1
R1 + R2
−
R1 R2
( R1 + R2 )
2
}* ud ) 2 ]1/ 2
R2
Case 2 provides the smaller uncertainty at the design-stage. We should proceed using this
design.
COMMENT
This is a classic illustration of using uncertainty analysis to determine the better of differing
approaches.
Although each individual resistor in Case 2 has a larger absolute uncertainty than those in
Case 1, we find that the weighted combination of the two resistors in Case 2 yields a
substantially lower uncertainty. This is not obvious by inspection alone. Our design results
from a close analysis of the sensitivity of the resultant to each contributing uncertainty. The
combination in Case 2 is just less sensitive to the individual uncertainties.
PROBLEM 5.10
KNOWN: p = 20 kPa @ FSO (see Note below)
(uo)3 = 0.01 kPa
(uo)4 = 0.001 kPa
FIND: Select 3 1/2 or 4 1/2 digit display based on uncertainty
SOLUTION
A design-stage analysis is appropriate here as the selection pertains to identical
instruments having only different output resolutions. At full-scale, the units will indicate
19.99 (3½) or 19.999 (4½ )
The instrument uncertainty is given by:
e1 = 0.0015 x 20 kPa = 0.03 kPa
e2 = .002 x 20 kPa = 0.04 kPa
e3 = 0.0025 x 20 kPa = 0.05 kPa
So that: uc = ( .032 + .042 + .052)1/2= 0.071 kPa
At FSO, the readout will display 19.99 or 19.999:
3 1/2 digit: ud= (.072 + .012)1/2 = 0.07 kPa
4 1/2 digit: ud = (.072 + .0012)1/2 = 0.071 kPa
The uncertainty is virtually identical regardless of the resolution in the readout. Meter
resolution does not affect the uncertainty to any practical extent.
Note: A true problem found several times in catalog of a major supplier of engineering
sensors and readouts in US.
PROBLEM 5.11
KNOWN: G = f(L,T,R, θ )
uL/L = uT/T = uR/R = u θ / θ = 0.01
FIND: (ud)G
SOLUTION
The shear modulus is found by G
= 2LT / πR 4 θ using the best estimates of L, T, R, and
θ . Its uncertainty is evaluated by
(
) (
) (
) (
)
1/ 2
2
2
2
2
uG / G =
±  u L / L + u T / T + 4u R / R + u θ / θ  = ±0.04 =
4%


Note that even if uL = uT = u θ = 0, the uncertainty uG/G is dominated by the uncertainty in R.
The shear modulus uncertainty is most ‘sensitive’ to the radius uncertainty.
PROBLEM 5.12
KNOWN: η = f(Tc , Th)
u η / η ≤ 0.01
Th = 40oC = 313 K
Tc = 20oC = 293 K
FIND: uh, uc required
SOLUTION
We will assume that the uncertainties in measuring either temperature are the same (i.e.,
u=
u=
u Th ) . Note that temperatures must be in absolute and that a 1 oC change equals a
T
Tc
1 K change.
1/ 2
2
2

  ∂η
 
∂η

± 
uη =
u  +
u  
 ∂T Tc   ∂T Th  
 
  h
 c
(
=
±  u Tc / Th

) (
2
)
1/ 2
2
+ u Th Tc / Th2 

Then, u η / η ≤ 0.01 requires: u=
( u=
u Th assumed) ≤ 0.1 K or 0.1 oC
T
Tc
In a laboratory environment, this low magnitude of uncertainty value in a measured
temperature would be attainable but with considerable care and calibration. In most
engineering processing plant applications, this value would be very difficult to attain.
PROBLEM 5.13
KNOWN: Heat transfer from a rod is to be determined.
Nu = hD/k is the nondimensional heat transfer.
uh = 150 ± 7% W/m2-K (95%)
ud = 20 ± 0.5 mm (95% assumed)
uk = 0.6 ± 2% W/m-K (95% assumed)
FIND: uNu
SOLUTION
At the known level of uncertainty provided: Nu = f(h,D,k) then,
1/ 2
u Nu
2
2
 ∂Nu 2  ∂Nu
  ∂Nu  
=
± 
uh  + 
u  +
u  
  ∂D D   ∂k k  
 ∂h
u Nu
2
2
 D  2  h
  −hD  
=
±  u h  +  u D  + 
uk  
  k2
 
 k   k
u Nu
2
2
2
 0.02
  150
  −150x0.02
 
=
± 
10.5  + 
0.005  + 
0.012  
0.62
  0.6
 
 
 0.6
=
1/ 2
=
1/ 2
=
±0.4
where
uh = (0.07)(150) = 10.5 W/m2-K
uD = 0.0005 m
uk = (0.02)(0.6) = 0.012 W/m-K
Then, we estimate the Nusselt number here to be
Nu = hD/k ± uNu = 5 ± 0.4 (95%)
So, Nusselt number can be determined within about 8% in this range of values.
COMMENT
The level of uncertainty analysis (design-stage, advanced design-stage, …) in the result
depends on the level of uncertainty values given or available. This problem solution provides
the propagation of uncertainty from the variables to the result.
PROBLEM 5.14
KNOWN: R = 30 Ω
P = 500 W
Ohmmeter
Resolution: 1 Ω
Accuracy: within 5% of reading
Ammeter
Resolution: 100 mA
Accuracy: within 0.1% of reading
FIND: (ud)E
SOLUTION
From Ohm's Law: E = IR or in terms of power, P =I2R. For the nominal values of power and
resistance given, expect a current, I = (P/R)1/2 = 4.08 A. Hence,
Ohmeter
(ud)R= (uo2 + uc2)1/2
= ((0.005 x 30 Ω )2+ (0.5 Ω )2)1/2
= 0.52 Ω
(95%)
Ammeter
(ud)I = (uo2 + uc2)1/2
= ((50 x 10-3 A)2+ (0.001 x 4.08 A)2)1/2
= 50.2 x 10-3 A
(95%)
Then, since voltage E = f(I,R):
1/ 2
2
2
 ∂E
  ∂E
 
± 
(u d ) E =
(u d ) I  + 
(u d ) R  
  ∂R
 
 ∂I
(
= ±  R(u d ) I

1/ 2
2
2
= ± ( 30Ω × 0.0502A ) + ( 4.08A × 0.52Ω ) 


) (
2
+ I(u d ) R
) 
2 1/ 2
= ± 2.61 V (95%)
COMMENT
Compare the groups in the (u d ) E term:
∂E
∂E
Current:
(u ) = 1.51 V
(u d ) R = 2.05V
∂I d I
∂R
Hence, the resistance measuring device contributes most to the uncertainty in voltage
measurement at the design stage. Focusing on reducing the uncertainty in the resistance
measurement would be a good starting place to reduce (u d ) E .
Resistance:
Note that the units in each of the working equations are consistent. The equations would not
be logical otherwise.
PROBLEM 5.15
SOLUTION
Design-Stage Analysis: Provides a quick (but not accurate) estimate in uncertainty based on a
planned approach. In general, this analysis is performed at a time when only information
about the uncertainies in measuring equipment and appropriate engineering constants
required for analysis are known or estimated. The analysis assumes perfect control of the
measurement process and its procedure.
Advanced-Stage Analysis: Provides an accurate estimate of the uncertainty in a result based
on a detailed knowledge of the measurement process. Used when only a single (or very few)
measurement(s) is planned. Such an analysis builds on a design-stage analysis estimate to
include estimates of uncertainty, such as procedural control, setability of operating
conditions, and repeatability of the measured variable.
PROBLEM 5.16
SOLUTION
Replication provides a measure of the control of the operating conditions and test procedure.
It does this by permitting the test engineer to quantify the differences in test results obtained
from distinctly duplicate tests conducted under nominally identical conditions. For example,
in a typical university lab course, several groups might setup and perform the same lab
exercise – but with different people at different times, etc. – these duplicate tests are
replications. Note: Often, a special type of replication, one in which the same test is
performed at a different test facility is referred to as a reproducibility test).
Repetition provides a measure of repeatability during the same test, so that the subtle
differences between test conduct (such as in a replication) are not included. For example, in a
typical university lab course, a group might take multiple data points at the same operating
condition from a test – these data points are repetitions.
In an advanced-stage analysis, replication effects are included as a higher-order uncertainty
based on estimates found by a trial of tests designed to measure such controllability. In a
multiple-measurement analysis, replication effects are usually entered as a precision error
evaluated from pooled statistics analysis.
PROBLEM 5.17
KNOWN: Displacement Transducer Instrument specifications
Linearity:
e1 = ± 0.25% reading
Drift:
e2 = ± 0.05%/oC reading
Sensitivity: e3 = ± 0.25% reading
Output Device specifications
Resolution: 10 µ V
Accuracy:
within ± 0.1% reading
o
Expect a 10 C variation during measurements.
Expect a nominal displacement of x = 2 cm.
FIND: (ud)x
SOLUTION
ud =
±[ud2 + ud2 ] 1/ 2 so find design-stage uncertainty in transducer and voltmeter:
x
T
E
Displacement Transducer
From the instrument specifications, we can assume that the static sensitivity of the
displacement transducer is K = 5 V/5 cm = 1 V/cm. Then combining the elemental errors for
this transducer and expecting a displacement of x = 2 cm and a temperature variation of up to
10C:
ud =
±[uo2 + uc2 ]T 1/ 2
T
where
uo = 0
(not relevant)
T
1/ 2
u c = e12 + e22 + e32 
T
= [(0.0025x22 + (.005 x 10oC x 2)2 + (0.0025 x 2)2 ]1/2
= 0.01225 V
u d = 0.01225 V
T
Voltmeter
ud =
±[uo2 + uc2 ]E1/ 2
E
where
u o = 5 x 10-6 V
E
u c = (2V x .001) = 0.002 V
E∂
u d = 0.002 V
E
Then, the design-stage uncertainty in measurement system becomes:
ud =
±[ud2 + ud2 ] 1/ 2 = ± 0.0124 V or (since K = 1 V/cm) = ± 0.0124 cm (95%)
x
T
E
COMMENT
Comparing each term in ud , we see that the transducer contributes most to the uncertainty
x
in measured displacement at the design stage. Keep in mind that at this level of uncertainty
only the intrinsic instrument errors and no procedural errors are considered.
PROBLEM 5.18
KNOWN: Measurement system of Problem 5.17
N = 20
x = 17.2 mm = 1.72 cm
Sx = 1.7 mm = 0.17 cm
FIND: x'
ASSUMPTIONS: Measurement is sufficiently controlled such that all errors have been
randomized and considered in the measured data.
SOLUTION
The measurements have provided additional information about the uncertainty involved in
the using this measurement system and involved in measuring this particular variable. We
will approach the problem as a multiple measurement uncertainty analysis problem. Using
the procedure of problem 5.17 but with x = 1.72 cm (recall Problem 5.17 used 2.0 cm),
u d = ± 0.0105 cm (95%)
T
u d = ± 0.0017 cm (95%)
E
Then, we can identify two elements of systematic error at the data acquisition source (see
Table 5.2) error due to the transducer, B2 , and error due to the output device, B4 (Note: the
subscripts used refer to the order in which these particular errors are listed in Table 5.2 – we
could easily just list them as known error 1 and known error 2).
B2 = u d = ± 0.0105 cm
T
B4 = u d = ± 0.0017 cm
E
We can set the random error in both of these elements to zero (no data provided).
P2 = P4 = 0
The twenty measurements are assumed to be made in a manner that best randomizes the
extraneous effects on the measured variable. The measurements provide a set of finite
statistics from which a precision interval in estimating the true mean value can be made.
Begin by estimating the random error due to temporal variation in the measurand:
P=
S=
S x / N = 0.038 cm
9
x
with degrees of freedom, ν = 19.
The measurement systematic error can be expressed as:
1/ 2
=
B  B22 + B42 
= 0.0108 cm
and the measurement standard random error is given by
P = 0.038 cm
The uncertainty interval is found from
(
)
1/ 2
2
ux =
±  B 2 + t19,95 P 


= ± 0.08 cm (95%)
with t19,95 = 2.093. The best estimate for mean mass displacement is
x' = 1.72 ± 0.08 cm
(95%)
based on the information provided.
COMMENT
Note that we could use this information as the basis for a single sample uncertainty
estimate by using the test performance obtained at 17.2 mm as representative for the
measurement system and procedure used and variable measured. This information could be
used to update the design stage analysis in the previous Problem.
PROBLEM 5.19
KNOWN: Nominal pressure value to be measured is 100 psi at 70oF
Pressure transducer
Accuracy: within 0.5% reading
Output device
Resolution:
0.1 psi
Sensitivity:
e1 = 0.1 psi
Linearity:
e2 = 0.1% reading
Drift:
e3 = 0.1 psi/6 months provided 32 < T < 90oF
ASSUMPTIONS: K = 1 V/psi for the transducer
FIND: ud
P
SOLUTION
For the transducer:
ud =
±[uo2 + uc2 ]T 1/ 2 = uc = ± 0.005 x 100 psi = ± 0.5 psi
T
For the output device
ud =
±[uo2 + uc2 ]D1/ 2 = ± 0.18 psi
D
where
u0 = ± 0.05 psi
1/ 2
uc = ± (e1 ) 2 + (e2 ) 2 + (e3 ) 2 
= ± 0.17 psi
= ± (0.12 + .12 + 0.12)1/2
Note: we have assumed a value for drift equivalent to that expected over the first 6 months of
operation following calibration. This value would be adjusted accordingly in an actual
situation. Then,
1/ 2
(95%)
ud =
± ud2 + ud2  = ± 0.53 psi
P
D 
 T
We anticipate an uncertainty of 0.53% of the reading at 100 psi due to the instrument errors
alone.
PROBLEM 5.20
KNOWN: p = 8610 lb/ft2
Sp = 273.1 lb/ft2
Np = 10
Bp = 1% rdg
D = 6.1 in. = 0.508ft
SD = 0.18 in.
ND = 10
BD = ± 1% rdg
t = 0.22 in.= 0.018 ft
St = 0.04 in.
Nt = 10
Bt = ± 1% rdg
FIND: σ '
SOLUTION
σ = σ ( p, D, t )
Pp = S p / N p = 86.4 lb/ft2
PD = S D / N D = 0.06 in = 0.005 ft;
Pt = St / N t = 0.01 in = 0.00105 ft.
1/ 2
2
2
2
 ∂σ
  ∂σ
  ∂σ  
Pσ = 
Pp  + 
PD  + 
Pt  
 ∂p
  ∂D   ∂t  
1/ 2
= 1197 2 + 1127 2 + 7087 2 
where p = p , D = D , t = t .
1/ 2
2
2
 D  2  p
  − pD  
=  Pp  +  PD  +  2 Pt  
  2t
 
 2t   2t
= 7275 lb/ft2
The value of ν σ is obtained from the Welch-Satterthwaite relation for a result that is a
function of variables:
2
( ) 
( ) ν
L
 ∑ θ i Px i
i =1
ν R = L
4
∑ θ i Px
i =1
i
2
2
xi
2
2
 D  2  p
  pD  
 Pp  +  PD  +  2 Pt  
  2t
 
 2t   2t
=
= 10
4
4
4
D 
 p

 Dp 
Pt 
 Pp 
 PD 

 2t  +  2t
 +  2t 2 
νp
νD
νt
1/ 2
2
2
2
 ∂σ
  ∂σ
  ∂σ  
Bσ = 
Bp  + 
BD  + 
Bt  
  ∂t
 
 ∂p
  ∂D
With Bp = 0.01 x 8610 = 86.1 lb/ft2 ,
1/ 2
2
2
2
 D
  p
  − pD  
=  B p  +  BD  +  2 Bt  
  2t
  2t
 
 2t
BD = 0.01 x 6.1 = 0.061 in = 0.005 ft ,
Bt = 0.01 x 0.22 = 0.0022 in = 0.0002 ft,
and p = p , D = D , t = t
1/ 2
Bσ = 11942 + 11742 + 13492  = 2150 lb/ft2
Then, with t10,95 = 2.228 and σ = pD / 2 t :
1/ 2
σ'=
σ ±  B 2 + (tP)2 
1/ 2
121, 497 ±  21502 + (2.238 × 7275) 2 
=
= 121,500 ± 16, 425 lb/ft2 (95%)
= 5.82 ± 0.786 MPa (95%)
PROBLEM 5.21
KNOWN: Calibration source elemental errors, K = 3
FIND: Random uncertainty due to calibration errors
SOLUTION
For three known random errors
1/ 2
P =  P12 + P22 + P32 
= 1.424 N/m2
= [0.92 + 1.12 + 0.092]1/2
with degrees of freedom,
2
 3 2
 ∑ Pk 
(0.92 + 1.12 + 0.092 ) 2
k =1

ν =
=
≈ 23
3
0.94 1.14 0.094
Pk4
+
+
∑ νk
20
9
14
k =1
PROBLEM 5.22
KNOWN: Systematic and random source errors in a measurement of force.
F = 200 N
FIND: Estimate F'
SOLUTION
The source errors can be combined to find the measurement systematic and random
uncertainties. The measurement systematic uncertainty is given by
1/ 2
B =  B12 + B22 + B32 
1/ 2
=  22 + 4.52 + 3.62 
= 6.1 N
Likewise, the measurement standard random uncertainty is given by
1/ 2
P =  P12 + P22 + P32 
1/ 2
= 02 + 6.12 + 4.22 
= 7.41 N
The degrees of freedom in the measurement standard random uncertainty is,
2
 3 2
 ∑ Pk 
(02 + 6.12 + 4.22 ) 2
k =1

ν =
=
≈ 32
3
6.14 4.24
Pk4
+
∑ νk
17
19
k =1
Then, using t32,95 = 2.042 (note: for ν ≥ 30 a value of t95 = 2 is commonly used as an
engineering approximation)
1/ 2
F ' = F ±  B 2 + (tP) 2 
F' = 200 ± 16 N (95%)
1/ 2
= 200 ± 6.12 + (2 × 7.4) 2 
N
PROBLEM 5.23
KNOWN: A = XY
X and Y Instrument Accuracy: within 0.5% reading
FIND: Estimate the uncertainty in land area
SOLUTION
The land area is found by A = XY. Then, the most probable estimate of the mean area is:
A = XY = 556 x 222 = 123,432 m2
In the measurement of length X (note: the subscripts refer to the order in which the particular
errors are listed in Table 5.2 – otherwise, they offer no special meaning):
Bx = B2 = 0.005 x 556 = 2.8 m
Px = P9 = Sx/N1/2 = 1.9 m
In the measurement of length Y:
By = B2 = 0.005 x 222 = 1.1 m
Py = P9 = Sx/N1/2 = 0.8 m
The propagation of these errors to the resultant area is found by
1/ 2
 ∂A 2  ∂A 2 
BA =
Bx  + 
By  
± 
  ∂Y
 
 ∂X
(
2
=
± (YBx ) + XBy

1/ 2
) 
2 1/ 2
= 873 m2
X =X
Y =Y
 ∂A 2  ∂A 2 
2 1/ 2
2
= 606 m2
PA =
Bx  + 
By   =
± 
± (YPx ) + XPy 
X
X
=
X
Y
∂
∂


 
 

Y =Y
The degrees of freedom in PA is found using a Welch-Sattertwaite relation
2
( )
L
2
2 2
2

θ
P
(
)
∑ i i 
(YPx ) + XPy 
i =1


νR = L
=
≈ 14
4
4
4
XP
∑ ( θi Pi ) ν xi (YPx ) + y
i =1
( )
( )
νx
νy
Then, with t14,95 = 2.145,
1/ 2
A' =
A ±  B 2 + (tP) 2 
1/ 2
=
123, 432 ± 8732 + (2.145 × 569) 2 
= 123,432 ± 1500 m2 (95%)
COMMENT
This uncertainty is about 1.75% of the measured area. The contribution from the instrument
itself is barely 0.7% of area. If we assume that the measurand does not change during
measurement (so barring seismic activity somewhat safely), the rest is due to lack of control
in procedure.
PROBLEM 5.24
KNOWN: σ = 1061 kPa
Sσ = 22 kPa
N = 23
FIND: P (listed as P due to temporal variation in Table 5.2)
ASSUMPTIONS: Scatter is due to temporal variations
SOLUTION
The standard random uncertainty, P, in the mean value of stress, σ , due to random error
by data scatter, Sσ , is
=
P S=
/ N 22 / 23 = 4.6 kPa
σ
with ν = N-1 = 22 degrees of freedom.
PROBLEM 5.25
KNOWN: N = 6 with Sx = 1.23 MPa
Bx= 1.48 MPa
FIND: ux
SOLUTION
The uncertainty in the mean value of strength is given by
1/ 2
ux =
±  Bx2 + (tPx ) 2 
where P=
S=
S x / N is the standard random uncertainty.
x
x
For t5,95 = 2.571 and S x = 1.23 MPa/61/2 = 0.50 MPa
1/ 2
ux =
± 1.482 + (2.571 × 0.5) 2 
=
± 1.96 MPa
(95%)
Both systematic and random errors contribute about the same.
PROBLEM 5.26
KNOWN: Standard: B1 = ± 0.5 psi P1 = 0
Voltmeter: B2 = ± 10 µV
P2 = 0
Measurement: B3 = ± .5 psi
Calibration Curve fit: P4 = Syx = 0.746 based on ν = 4
FIND: up
SOLUTION
From the calibration data, a least squares fit yields:
y =ao + a1 x ± tS yx or p [psi] =0.54 + 24.03E [mV] ± (2.776)(.746)
so that,
P4 = Syx = 0.746
based on ν = 4. Because P4 is the only random error involved, the measurement standard
random uncertainty is P = P4 = 0.746.
Also, from the curve we find that K =
∂p / ∂E = 24.03 psi/mV. We use this sensitivity to
convert between psi and mV.
1/ 2
B=
±  B12 + B22 + B32 
1/ 2
=
± (0.5 psi ) 2 + (0.01mV × 24.03 psi / mV ) 2 + (0.5 psi ) 2 
= ± 0.707 psi
Then, with t4,95 = 2.776, the combined uncertainty is
1/ 2
up =
±  B p2 + (tPp ) 2 


1/ 2
=
± 0.707 2 + (2.776 × 0.746) 2 
=
±2.19 psi (95%)
PROBLEM 5.27
KNOWN: Density of metal composite is determined by mass estimation.
Sample ingot is cylindrical in shape.
Nominal* values for typical ingot:
m ≈ 4.5 lbm
uo = 0.1 lbm
m
L ≈ 6 in.
uo = 0.05 in.
D ≈ 4 in.
uo = 0.0005 in.
L
D
uc / m = 1% reading; uc / L = 1% reading; ucD / D = 1% reading
m
*
L
"nominal" means 'approximate' or 'typical' values for a given situation
FIND: ud
ρ
SOLUTION
= 4m / π D 2 L
∀ =π D 2 L / 4 and ρ= m / ∀
1/ 2
2
2
2
 ∂ρ
  ∂ρ
  ∂ρ  
± 
uρ =
um  + 
uD  + 
uL  
 ∂m   ∂D   ∂L  
2
2
2 1/ 2
 4
  −8m
  4m
 
um  + 
uD  + 
uL  
=
± 
2
3
2
 π D L   π D L   π D L  
1/ 2
=[0.12 + (0.01 × 4.5) 2 ]1/ 2 = 0.11 lbm
1/ 2
=[0.0052 + (0.01 × 4)2 ]1/ 2 = 0.04 in = 0.0034 ft
1/ 2
=[0.052 + (0.01 × 6) 2 ]1/ 2 = 0.078 in = 0.0065 ft
um =uo2 + uc2 
u D =uo2 + uc2 
u L =uo2 + uc2 
1/ 2
2
2
2
uρ =
± ( 0.0133 × 0.11) + ( 0.0298 × 0.04 ) + ( 0.0597 × 0.078 ) 


3
3
= ± 0.0050 lbm /in = ± 8.680 lbm/ft
The mass measurement contributes most to the uncertainty at the design stage (= 0.0133 x
0.11), although just more than the diameter measurement, and deserves first attention to
reducing the uncertainty in density. The uncertainty contribution due to length ( = 0.005 x
0.078) is one order of magnitude smaller than mass or diameter.
PROBLEM 5.28
KNOWN: M = 3 replications with N = 10 repetitions each.
Sample mean and sample standard deviation values.
FIND:
d ± ud
SOLUTION
We will assume that errors enter only at the data acquisition stage. Elemental errors from
data acquisition sources will consist at least as a systematic error due to instrument error (B1),
a random error due to the variation in readings measured at each cross-section (P2), and
spatial-dependent random errors along the ingot length (P3).
The pooled mean diameter is found by
1
d =
(3.992 + 3.9892 + 3.9961) = 3.9924 inches
3
and provides the most probable estimate in the true mean diameter. The variation in readings
taken at each cross-section location is estimated by the pooled standard deviation relative to
the pooled mean. For N = 10 at M =3,
=
Sd = Sd / MN
1
(0.0052 + 0.0012 + 0.00092 ) / 3 × 10
3
= 0.00055 inches
with ν =M ( N − 1) =3 × (10 − 1) = 27. This yields a measure of the standard random error
due to data scatter P2.
The spatial variation in mean values is estimated by the standard deviation
1/ 2
M
2
 ∑ (d m − d ) 

Sd =  m=1


M −1




= 0.0034 inches
with ν = M - 1 = 2. The standard random error due to spatial errors P3 is then
P3 = Sd / M = 0.0017 in.
Then,
From the previous problem, the instrument error in diameter measurement is 1% of the
reading. So
B = B2 = Bc = 0.040 in.
D
1/ 2
1/ 2
P =  P2 + P32  = 0.000552 + 0.0017 2 
2
= 0.0018 in.
with
2
 2 2
 ∑ Pk 
(0.000552 + 0.0017 2 ) 2
k =1


=
=
= 2.5 ≈ 2
ν
2
0.000554 0.0017 4
Pk4
+
∑ νk
27
2
k =1
so t2,95 = 4.303 (note: ν gets rounded down).
1/ 2
ud =
±  B 2 + (tP) 2 
B = 0.040 in.
= ± 0.041 in.
P = 0.0018 in. ν = 2
d' = 3.9924 ± 0.041
in. (95%)
COMMENT
The large systematic error in the diameter measuring device dominates this problem. Even
though all the readings appear to be well behaved with some variations in position and
between positions, the systematic error implies these readings may all have an offset.
This problem also provides a nice example of how the statement of a value of a variable such
as the diameter of an ingot or of a shaft is actually a statistical statement. We often think of
such non-temporal variables as being fixed and absolute. Because they are statistical
statements, there is an associated uncertainty in their values.
PROBLEM 5.29
KNOWN: Measurements of mass and length and results of Problem 5.28
FIND: Estimate ρ '
SOLUTION
We will assume that errors enter only at the data acquisition stage.
Elemental errors from data acquisition sources will consist at least as a systematic error
due to instrument error (B1), a random error due to the scatter in data (P2). We will also use
the systematic and random error estimates from the diameter measurement information in
Problem 5.28. These give:
Bd = 0.040 inch, Pd = 0.0018 inch, ν = 2
The instrument error is estimated at 1% of the reading for each variable.
Bm = B2m = 0.045 lbm
BL = B2L = 0.06 inch
The scatter in the data creates a random error in estimating the mean value. The random error
estimates in the mean values are assumed as
Pm = P2m = Sm/N1/2 = 0.1/211/2 = 0.022 lbm
ν m = 20
PL = P2L = SL/N1/2 = 0.1/111/2 = 0.0302 in. ν L = 10
With ρ = 4m / π D 2 L ,
1/ 2
2
2
2
 ∂ρ
  ∂ρ
  ∂ρ
 
± 
Bρ =
Bm  + 
BD  + 
BL  
  ∂D
  ∂L
 
 ∂m
2
2
2 1/ 2
 4
  −8m
  4m
 
Bm  + 
BD  + 
BL  
=
± 
2
  π D3 L
  π D 2 L  
 π D L
1/ 2
2
2
2
 ∂ρ
  ∂ρ
  ∂ρ  
Pρ =
Pm  + 
PD  + 
PL  
± 
 ∂m   ∂D   ∂L  
2
2
2 1/ 2
 4
  −8m
  4m
 
Pm  + 
PD  + 
PL  
=
± 
2
3
2
 π D L   π D L   π D L  
Subbing:
1/ 2
2
2
2

−8 × 4.4
4
4 × 4.4
 
 
 
± 
Bρ =
0.045  + 
0.040  + 
0.06  
2
  π × 3.99243 × 5.85
  π × 3.99242 × 5.85
 
 π × 3.9924 × 5.85
Bρ = ± = 0.00385 lbm/in3
1/ 2
2
2
2

−8 × 4.4
4
4 × 4.4
 
 
 
± 
Pρ =
0.022  + 
0.0018  + 
0.032  
2
  π × 3.99243 × 5.85
  π × 3.99242 × 5.85
 
 π × 3.9924 × 5.85
Pρ = ± 0.00194lbm/in3
where each term is evaluated at the mean values for m, L and d.
νR
=
L
2
∑ ( θi Pi ) 
 i =1

L
∑ ( θi Pi )
i =1
4
2
νx
≈ 22
i
1/ 2
so that with u ρ =
±  B 2 + (tP) 2 
with t22,95 = 2.07
1/ 2
uρ =
± 0.003852 + (2.07 × 0.00194) 2 
ρ = 4m / π D 2 L
=
±0.0056 lbm/in3 (95%)
= 0.0596 lbm/in3
Then,
ρ ' = 0.0596 ± 0.0056 lbm/in3 (95%)
The updated value here contains additional information relative to a design stage analysis. In
this updated analysis, the random errors are based on actual measured data scatter and are
larger than the previous estimate that was based on resolution error (uo) alone. The
systematic errors are the same as the uc values used previously.
PROBLEM 5.30
KNOWN: Calibration against a standard.
Calibration data are provided.
FIND: a.) Calibration curve fit. b.) Uncertainty in any estimated T.
SOLUTION
a.) To compute a calibration curve fit, the data are fit to a polynomial using a least squares
analysis (spreadsheet programs have this capability). An acceptable first order curve fit of the
form y =a o + a1x ± tSyx is found to be:
E [mV] = -0.0219 + 0.0416T[oC] ± 0.086 mV (95%) with t4,95 = 2.770 and Syx = 0.031mV
This curve has a static sensitivity of 0.0416mV/oC. It corresponds to the curve:
T[oC] = 0.540 + 24.03E[mV] ± 2.08oC (95%) with Syx = 0.75oC
which has a static sensitivity of 24.03oC/mV. We see that a very small change in voltage
corresponds to a large change in temperature! This is typical of thermocouples.
b.) From the problem statement, we can identify two errors from the calibration source:
B1 = 0.05oC
B2 = 5oC/m x 0.010m = 0.05oC
Each value of independent variable is measured once and the results combined to form the
curve fit in part a. Voltage measurement system systematic errors are not indicated
but will be present in the measured data. Lacking other information, we will make the
assumption that these errors are of the magnitude of the resolution of the instrument, so that
we set:
B3 = 0.001mV = 0.024oC
(Note: from experience or manufacturer specifications, we might change this value up or
down – but we leave as is here).
Random errors in the measurement system are presumed included in the calibration curve.
The random error in the calibration curve is estimated as
P4 = Syx = 0.75oC with ν = 4.
So,
1/ 2
B =  B12 + B22 + B32  =
P = 0.75OC ν = 4
[0.052 + 0.052 + 0.0242 ]1/ 2 =
0.075o C
1/ 2
±2.08o C (95%)
uT =
±  B 2 + (tP ) 2  = ± [ 0.0752 + (2.77 *0.75) 2 ] =
Here the data scatter on the curve fit has the greatest effect on the uncertainty. In practice,
systematic errors can be more than an order of magnitude higher than used here.
1/ 2
PROBLEM 5.31
KNOWN: P = E2/R
P ≈ 100 W (nominal values)
R ≈ 100Ω
Ohmmeter
Resolution: 1 Ω
Error:
1% reading
Voltmeter
Resolution: 1V
Error:
1% reading
FIND: uo and ud for power
SOLUTION
For a nominal power of 100 W, we get a nominal value for voltage:
=
E =
PR
(100Ω)(100W) = 100 V
Uncertainty in power is due to uncertainties in E and R:
1/ 2
1/ 2
2
2
 ∂P 2  ∂P
 2E 2  −E 2
 
 
(1)
± 
± 
uP =
uE  + 
uR   =
uE  + 
u  
 
  R 2 R  
 ∂E   ∂R
 R
To find uncertainty at any order (e.g., uo or ud) , use the uncertainties at that order and
evaluate (1) using E = 100 V and =
R 100Ω .
Zero-order uncertainty
u o = ± 0.5V
uo =
±0.5Ω
E
R
so substituting these values into (1), u o = ± 1.12 W
(95%) .
p
Design-stage uncertainty
1/ 2
=
u d  u o2 + u c2 
The estimates for uo are given above. The estimates for uc are
u d =± (100Ω)(0.01) =±1Ω
ud =
±(100V)(.01) =
± 1V
E
R
so,
u d = ± 1.12V
E
ud =
±1.12Ω
R
Substituting these values into (1):
u d = ±2.5W (95%)
P
The uncertainty estimate becomes more accurate as each new piece of information is added.
PROBLEM 5.32
KNOWN: P = 10, 1000, 10000 W
P = E2/R or P = IE
Instrument specifications in Table
FIND: Design (select) a best method using uncertainty analysis
SOLUTION
This problem is open-ended and this solution is offered as a guide.
Suppose we fix E = 100V. This determines R and I for analysis. With the information
available, a design-stage analysis is possible.
For P=f(E,R):
1/ 2
1/ 2
2
2
 ∂P 2  ∂P
 2E 2  −E 2
 
 
± 
± 
uP =
uE  + 
uR   =
uE  + 
u  
 
  R 2 R  
 ∂E   ∂R
 R
or in terms of fractional percent, the relative uncertainty is (divide through by P=E2/R )
(
uP / P =
±  2u E / E

For P=f(I,E)
) (
2
)
1/ 2
2
+ uR / R 

(1)
1/ 2
 ∂P 2  ∂P 2 
2
2 1/ 2
uP =
± 
uE  +  uI   =
±  Iu E + Eu I 


 ∂E   ∂I  
or, the relative uncertainty (divide through by P = EI)
(
uP / P =
±  uE / E

) + ( u / I ) 
2 1/ 2
2
(2)
I
1/ 2
=
u d  u o2 + u c2 
= [0.52 + (0.005 x E)2]1/2
At the design stage:
ud
( ) ( )
voltage
E
u d = [0.252 + (0.01 x A)2]1/2
current
u d = [0.52 + (0.005 x R)2]1/2
resistance
(3)
A
R
Method 1: Solve for P = E2/R , equation (3) and then equation (1)
P
E
R
uP /P
[W]
[V]
[Ω]
[%]
10
100
1000
1.5
100
100
10
2.9
1000
100
1
50.0
Method 2:
P
[W]
10
100
1000
Solve for P = EI , equation (3) and then equation (2)
E
I
uP/P
[V]
[A]
[%]
100
0.1
250
100
10
2.8
100
100
1.3
Method 1 is better at low power levels while method 2 is better at high levels.
COMMENT
Different values of E will produce different results. A broader look of this problem would
vary E and optimize to determine a basis for preferred operating conditions (in E, R, and I).
PROBLEM 5.33
KNOWN: Composite material is function of cure temperature, σ = f(T).
Possible cure temperature range: 20oC to 60oC
Oven controllability to be tested at 30oC:
Oven divided into quadrants: j = 1 to J
N measurements taken in each quadrant: i = 1 to N
M replications to be made of entire test: m = 1 to M
Method 1: N = 5, M = 5, J = 4 i.e. JxNxM = 100
Method 2: N = 25, M = 1, J = 4 i.e. JxNxM = 100
FIND: If uncertainty in the oven test temperature is to be estimated, discuss
information obtained by the two different methods.
ASSUMPTIONS: In order to simplify the solution we assume that sensor installation effects
and measurement system operating conditions are properly controlled. We also neglect data
reduction errors. In both cases the effects will be the same for either method.
SOLUTION
Open-ended problem. This is an excellent problem for an instructor-directed group
discussion.
Our goal is to estimate the uncertainty associated with the σ (T) test due to controllability of
the independent variable, T. This test is really one of oven performance. By setting the oven
to one representative condition, we can estimate typical oven performance at other
temperatures through a single measurement analysis.
At any set temperature:
(i) By dividing the oven into quadrants and determining the mean quadrant temperature,
we obtain information concerning the typical oven spatial variation in temperature.
(ii) By repetition, we obtain information about the typical oven temporal variation in
temperature.
(iii) By replication, we obtain information about our ability to repeat the exact conditions on
subsequent attempts (obviously, variations in the actual achieved oven temperature, despite
the fact that the oven controls are seemingly reset exactly the same on each replication,
affects strength).
Define:
u1: temporal variation effect on mean oven temperature
u2: spatial variation effect on mean oven temperature
u3: set controllability (repeatability) of the mean oven temperature
uc: instrument error associated with the measuring equipment
Mean temperature of any quadrant on any replication:
1 N
Tjm = ∑ Tijm
N i =1
Grand pooled mean temperature:
T =
1 M
∑T
M m =1 m
Pooled mean oven temperature on any replication:
1 J
Tm = ∑ Tjm
J j=1
Spatial temperature variation on any replication:
1/ 2
2
 1 J =4
ST
T
T
=
−

∑ jm m 
m
 J − 1 j=1

Temporal temperature variation on any replication:
)
(
J =4 N
 1
ST
Tijm − Tm
=

∑∑
m
 J(N − 1) =j 1 =i 1
Set controllability of mean temperature:
(
(
)
1/ 2
2
) 
1/ 2
2
 1 M
=
−
S
T
T


∑ m
T
 M − 1 m =1

The Nth order uncertainty is found from
1/ 2
uN =
±  u12 + u 22 + u 32 + u c2 
Method 1:
(P%)
Method 2:
u1 = t ST
/ JN
m
u1 = t ST
/ JN
m
u 2 = tST / J
u 2 = tST / J
u3 = t S
u3 = information not available without replication
m
m
/ M
T
Without replication we will have no information on our ability to control the oven set
temperature. This will be an important omission if we intend to use the oven for batch
production.
COMMENT
There may be other options as to how to approach this problem and the above analysis
presents but one approach. One alternative performs a multiple measurement analysis to
estimate our ability to control the oven at 30oC. Using,
P1 = ST
m
/ JN ; P2 = ST / J ; P3 = S
we obtain a similar result.
m
T
/ M ; B4 = uc
PROBLEM 5.34
KNOWN: N1 = 50
x = 2.112 V
S1= 0.387
CI ≤ 0.10 at 95%
FIND: NT
SOLUTION
The confidence interval is a two sided interval given by
CI = ± tS / N1/ 2
For a CI = 0.10, if d = CI/2 = tS / N1/ 2 , then d = 0.05. Evaluate an NT based on the
available S1. Then,
N T ≈ (t N −1,95S1 / d) 2 × 0.387 / 0.05) 2 = 242
PROBLEM 5.35
KNOWN: N = 10 measurements made at M = 4 locations
Micrometer: Resolution: 0.001 in.; Accuracy: < 0.001 in.
FIND: D'
ASSUMPTIONS: We will restrict the solution to errors due to data acquisition sources.
SOLUTION
A pooled estimate of the diameter of the shaft is given by
=
D'
4.494 + 4.499 + 4.511 + 4.522
= 4.506 in.
4
with standard deviation
SD =
0.0062 + 0.0092 + 0.012 + 0.0032
= 0.0075 in.
4
We can recognize elemental errors due to instrument error and due to spatial variations
effects on the computed mean value and procedural variation errors which bring about data
scatter at each cross section. Instrument errors will be treated as a systematic error,
B1 = uc = 0.001 in.
This value reflects the manufacturer accuracy statement. The procedural variations bring on a
random error in each measured mean value as estimated by
P2 = S
D
= <SD>/(MN)1/2= 0.0012 in.
with degrees of freedom, ν = M(Nj - 1). The spatial error arises because the diameter is not
uniform along its length such that the mean values vary. This spatial error affects the pooled
mean value and is estimated by
 M =4 (D − D
j
SD =  ∑
 m =1 M − 1

1/ 2




= 0.0126 in.
P3 = S = SD/M1/2 = 0.006 in.
D
Collecting random errors,
P = [P22 + P32]1/2 = 0.006 in.
with degrees of freedom from the Welch-Sattertwaite approach
=
νD
(0.0012 + 0.0062 ) 2
≈3
0.0014 0.0064
+
36
3
The uncertainty estimate can be given by
uD = [B2 + (t3,95P)2]1/2= [.0012 + (3.182 x 0.006)2]1/2
D' = 4.506 ± 0.019 in. (95%)
COMMENT
Some observations on how the Welch-Sattertwaite formula weights the degrees of freedom:
- in cases where one degree of freedom is substantially less than another while the
random errors are of the same order of magnitude, that smaller degree of freedom
dominates
in cases where one random error is substantially greater than the other while the degrees of
freedom are either small but not too different or all are large ( >30), the larger random error
value dominates.
PROBLEM 5.36
KNOWN: Pressure is measured using a dial gauge.
p = 50 psi Sp = 2 psi M = 30 N = 1
Dial gauge:
Resolution: 0.1 psi
Accuracy: within 0.5 psi
FIND: Estimate the uncertainty in vessel pressure.
SOLUTION
During a process the pressure is known to vary due to the set point of the compressor.
Pressure will be set at p and readings are to be taken. We are asked to estimate the
uncertainty in the set pressure at any measurement. We do this by running a trial set of data
for pressure set point versus actual process pressure:
Single measurement analysis: From the manufacturer statement, we set
uc = 0.5 psi
The variations in vessel pressure upon each replication are estimated by
Sp = 2 psi
such that,
u1 = t29,95Sp= (2.047)(2) = 4.1 psi
So if during a test we set the vessel pressure, then the uncertainty in the process pressure
during any one measurement is
uN = ± [ 0.52 + 4.12 ]
1/ 2
= ± 4.1 psi (95%)
COMMENT
On 30 trials, we found a standard deviation in the set point to be 2 psi. The uncertainty in the
set point during any one trial should be about twice this number for 95% probability. Notice
how this is quite different then trying to determine the average set point over 30 trials (which
works out to be only 0.9 psi at 95% - see below). This is an example of the advanced stage
analysis applied to access uncertainty in a single trial (sample).
The uncertainty in the actual mean set point is:
u1 = t29,95Sp/M1/2 = (2.047)(2)/301/2 = 0.75 psi
uN = ± [ 0.52 + 0.752 ]
1/ 2
= 0.91 psi (95%)
PROBLEM 5.37
KNOWN: Transducer, readout specifications.
M = 4 replications with N = 10 repetitions each.
FIND: x ± u x
SOLUTION
From the data, the pooled mean: x = (4.3 + 3.8 + 4.2 + 4.0)/4 = 4.08
At the design-stage, only information known prior to the test are included:
For the transducer, we estimate the instrument error from the elemental errors:
eL = ± (0.0025)(4m) = 0.01 m
eR = ± (0.0025)(4m) = 0.01 m
eK = ± (0.001)(5m) = 0.005 m
ez = ± (0.005/oC)(3oC)(5m) = 0.0075 m
so that,
uc = (.012 + .012 + .0052 + .00752)1/2 = 0.0167 m
t
The transducer has a sensitivity Kt = 1V/m so that the voltmeter output can be restated in
terms of displacement [m].
For the voltmeter:
uc = {[(0.001)(4m)(1m/V)]2 + [(5 µ V)(1m/V)]2}1/2 = 0.004 m
E
For the transducer-voltmeter system, the design-stage uncertainty is:
ud = ± (.0042 + .01672)1/2 = ± 0.017 m (95%)
We can use this information for the multiple-measurement analysis where we assign
instrument errors:
B1 = uc = 0.0167 m
B2 = uc = 0.004 m
t
E
The uncertainty in the applied input estimates the control of the input forcing function.
Assuming F = kx, then K = (2000 N)/(4 m) = 500 N/m. Hence, we assign:
1/ 2
2
 ∂x
 
B3 = 
u F   = 0.2 m
 
 ∂F
The temporal or repetition error in the mean values is estimated from the given deviations:
P4 = [(S12 + S22 + S32 + S42)/4]1/2/(MN)1/2 = 0.04 m with ν 4 = 36
The temporal error or replication error in the overall pooled mean is:
1/ 2
 4

2
P5 =
 ∑ ( x j − x ) /( M − 1) 
 m=1

0.11 m with ν 5 = 4
2
0.042 + 0.112 
 =4
ν=
4
4
0.04
0.11
+
36
4
with t4,95 = 2.770. Then, for the measurement:
B = (.01672 + .0042 + .22)1/2 = 0.2 m
So, the ability to control the applied force dominates the systematic error.
P = (.112 + .042)1/2 = 0.12 m
Then, ux = ± (.22 + [(2.770)(.12)]2)1/2 = ± 0.39 m (95%)
Hence,
x' = x ± u x = 4.08 ± 0.39 m (95%)
PROBLEM 5.38
KNOWN: First-order system:
uΓ/Γ = ± 0.02 (95%)
ut/t = ± 0.01 (95%)
FIND: uτ / τ
ASSUMPTION: τ= f (Γ, t ) with no other influences.
SOLUTION
Γ =e−t / τ
Rearranging, τ =
−t / ln Γ
∂τ
 ∂τ

uτ =
± ( ut ) 2 + ( uΓ ) 2 
∂Γ
 ∂t

in relative terms,
1/ 2
1/ 2
u
 u

± ( t )2 + ( Γ )2 
uτ / τ =
Γ(ln Γ) 
 t
Γ
.1
.5
.8
.9
1.0
uτ / τ
0.013
0.031
0.090
0.190
∞
1/ 2
uΓ
 u
2
)
=
± ( t )2 + (

Γ(ln Γ) 2 
 ln Γ
PROBLEM 5.39
KNOWN: Transducer, readout specifications.
M = 3 replications with N = 20 repetitions each.
FIND: T ± u T
SOLUTION
From the data, T = (181.0 + 183.1 + 182.1)/3 = 182.1 oC
From the problem statement, we can assign
B1 = e1 = 1 oC measuring system error
B2 = e2 = 1.2 oC installation error
The temporal or repetition error in the mean values is estimated from the variations in the
data sets:
P3 = [(S12+ S12 + S32)/3]1/2 /(MN)1/2 = 0.38 oC with ν3 = M(Nm-1) = 57
The temporal error or replication error in the overall pooled mean is:
1/ 2
 3

P4 =  ∑ (Tm − T ) 2  / M1/ 2 = 0.61 m with ν 4 = (M-1) = 2
 m =1

Then, for the measurement:
B = (12 + 1.22)1/2= 1.56 oC
P = (.382+ .612)1/2= 0.72 oC
with, using the Welch-Satterthwate estimate, ν= f (ν3 , ν 4 ) = 3.8 ≈ 4
and t4,95 = 2.770. Then,
uT = ± (1.562 + [(2.770)(.72)]2)1/2 = ± 2.5 oC (95%)
Hence,
T' = T ± u T= 182.1 ± 2.5o C (95%)
PROBLEM 5.40
SOLUTION
Exact answers will depend on user experience and specific instruments user has experience
with. As a class exercise, it would be interesting to discuss why different people chose their
numbers.
As a guide:
bathroom scale:
spring scale: ± 5 to 10 N
balance beam scale: ± 1 N
plastic ruler: to within its resolution
micrometer: to within its resolution
kitchen thermometer: ± 1 oC
speedometer: ± 4% reading (in USA this is by law)
PROBLEM 5.41
KNOWN: Air @ T = 25 ± 2 oC (95%)
up/p = 0.01 (95%)
FIND: u ρ / ρ
ASSUMPTIONS: Ideal gas p = ρ RT. Neglect uncertainty in gas constant R.
SOLUTION
ρ = f ( p, T )
1/ 2
u ρ / ρ (u p / p ) 2 + (uT / T ) 2 
=


A change in 1 oC equals a change of 1 K. So we can rewrite temperature as
T = (25 + 273) ± 2 K = 298 ± 2 K (95%)
1/ 2
(0.01) 2 + (2 K / 298K ) 2 
uρ / ρ =


=
±0.012 or ± 1.2% (95%)
PROBLEM 5.42
KNOWN: Air at T = 25oC with uT = ± 1 oC (95%)
u ρ / ρ = ± 0.005
FIND: up/p
ASSUMPTIONS: Ideal gas p = ρ RT. Neglect uncertainty in gas constant, R.
SOLUTION
1/ 2
± (u ρ / ρ ) 2 + (uT / T ) 2 
up / p =
A change in 1 oC equals a change of 1 K. So, T = 298 ± 1 K (95%). Subbing terms,
1/ 2
up / p =
± (0.005) 2 + (1K / 298 K ) 2 
=
±0.0037 or ± 3.7% (95%)
PROBLEM 4.43
KNOWN: T = e-KEW and 0 ≤ T ≤ 1
E = 2 ± 0.04 (95%)
uT/T = ± 0.01 (95%)
uW/W = ± 0.01 (95%)
FIND: Is u K / K ≤ 0.05? 0.10?
SOLUTION
The solids density is found by
K = - (ln T)/EW
so that,
1/ 2
uK / K =
± (uT / T ln T ) 2 + (u E / E ) 2 + (uW / W ) 2 
1/ 2
=
± (0.01/ ln T ) 2 + (0.04 / 2) 2 + (0.01) 2 
Then for values of transmission factor, T :
uK/K
.5
.8
.9
1.
0.027
0.05
0.0975
∞
So uK/K
approaches 5% at
T = 0.8 and 10%
at T = 0.9.
1.5
1
uK/K
T
0.5
0
0
0.2
0.4
0.6
T
We see that
u K / K → ∞ as T → 1 .
0.8
1
PROBLEM 5.44
KNOWN: First-order system:
uΓ/Γ = ± 0.02 (95%)
ut/t = ± 0.005 (95%)
FIND: uτ / τ
ASSUMPTION: τ= f (Γ, t ) with no other influences.
SOLUTION
Γ =e−t / τ
Rearranging, τ =
−t / ln Γ
∂τ
 ∂τ

uτ =
± ( ut ) 2 + ( uΓ ) 2 
∂Γ
 ∂t

in relative terms,
1/ 2
1/ 2
uΓ
 u
2
)
=
± ( t )2 + (

Γ(ln Γ) 2 
 ln Γ
1/ 2
u
 u

± ( t )2 + ( Γ )2 
uτ / τ =
Γ(ln Γ) 
 t
1/ 2
0.02 2 

=
± (0.005) 2 + (
)
ln Γ 

Γ
.1
.5
.8
.9
1.0
uτ / τ
0.010
0.029
0.090
0.190
∞
Because τ approaches steady state asymptotically as Γ goes to 1, this behavior
is to be expected.
PROBLEM 5.45
KNOWN: acceleration, a = f(L,s,t1,t2)
L = 5.0 ± 0.5 cm
s = 100.0 ± 0.2 cm
t1 = 0.054 ± 0.001s
t2 = 0.031 ± 0.001s
all uncertainties stated at 95% confidence
FIND: ua
SOLUTION
We know that
L2 1 1
=
a
( − )
2 s t22 t12
Applying the nominal values into the equation gives,
52
1
1
=
a
− =
(
) 87.21 cm/s2
2
2
2 × 100 0.0312 0.0541
And ua = f (u L , us , ut , ut ) . Expanding,
1
2
1/ 2
 ∂a

∂a
∂a
∂a
ua =
± ( u L ) 2 + ( us ) 2 + ( ut ) 2 + (
ut ) 2 
∂s
∂t1 1
∂t2 2 
 ∂L
where
∂a L2 1 1
∂a 2 L 1 1
= 2 ( 2 − 2)
=
( − )
∂s s t2 t1
∂L 2 s t22 t12
∂a L2 2
=
( )
∂t2 2 s t23
∂a L2
2
=
(− 3 )
∂t1 2s t1
∂a
5
1
1
=
uL
(
−
)(0.05)
2
∂L
100 0.031
0.0542
∂a
52
1
1
=
us
(
−
)(0.2)
2
2
∂s
100 0.031 0.0542
∂a
52
2
=
ut
(−
)(0.001)
∂t1 1 200 0.0543
∂a
52
2
ut =
(
)(0.001)
∂t2 2 200 0.0313
ua =
±(1.7442 + 1.7442 + 1.587 2 + 8.3922 )1/ 2 =
±8.89 cm/s2
and so the acceleration is best estimated as
a = 87.21 ± 8.89 cm/s2
(95%)
or ua/a is about 10.2%
As a concomitant check, we can estimate acceleration from a = gsin α . Note that if α is
measured, then acceleration a is easily anticipated. Of course, such an estimate neglects
errors due to systematic effects, such as friction. But friction effects are built into the first
estimate and the differences will reflect the magnitude of this systematic error. With
g = 9.806 m/s2 = 980.6 cm/s2, we assign an uncertainty of ug = ± 0.1 cm/s2 to this
gravitational constant (equal to its roundoff error).
For example, for an angle α =30. Then with,
1/ 2
ua (u g sin α ) 2 + ( guα cos α ) 2 
=


2
For ua ≤ 8.89 cm/s , we need uα ≤ 0.6o (=0.01047 rad).
As α increases, the uncertainty becomes less sensitive to the angle measurement and more
sensitive to gravity. Note that as α → 90o ,
1/ 2
(u sin α ) 2 + (0) 2  → u
u=
a
g
 g

and the uncertainty in acceleration becomes independent of the uncertainty in α .
PROBLEM 5.46
KNOWN:
Measured data relating golf ball carry distance to launch angle and initial
velocity
FIND:
Uncertainty in carry distance, D, as determined from initial velocity, Vo and
launch angle, φ
SOLUTION:
The uncertainty in distance is determined from
2
2
 ∂D
  ∂D 
u D = 
uV  + 
uφ 
 ∂Vo
  ∂φ 
with the partial derivatives approximated numerically from measured data. The units of angle
used are in the equations is radians.
o
Initial
Velocity
(MPH)
165.5
167.8
170
172.2
165.5
167.8
170
172.2
165.5
167.8
170
172.2
Launch Angle
(degrees)
8
8
8
8
10
10
10
10
12
12
12
12
Carry Distance
(yds)
254.6
258.0
261.4
264.8
258.2
261.6
264.7
267.9
260.6
263.7
266.8
269.8
∂D
∂Vo
1.48
1.55
1.55
1.55
1.48
1.41
1.45
1.45
1.48
1.41
1.45
1.45
∂D
∂φ
uD
1.8
1.8
1.7
1.6
1.2
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.49
1.55
1.56
1.56
1.49
1.42
1.45
1.45
1.48
1.41
1.45
1.45
So a single value of uncertainty, approximately 1.5 yds (1.37 m), can be used to quantify the
effects of errors in the measurement of Vo and φ . Note that the only significant contribution
to uncertainty results from uncertainty in the initial velocity.
PROBLEM 5.47
KNOWN:
Measured data relating pressure drop and volumetric flow rate
FIND:
Required uncertainty in pressure drop to yield a 0.25% uncertainty in
volumetric flow rate
ASSUMPTION:
For this problem, assume that the only errors enter in the measurement of
pressure drop. In practice, errors related to dimensions and hydraulic
coefficients will add additional errors.
SOLUTION:
The uncertainty in flow rate is
uQ
2
 ∂Q
 
= 
u ∆P  
 ∂∆P
 
1/ 2
or in relative terms we can reduce this to
uQ
Q
=
u ∆P
∆P
Q (m3/min)
∆P (Pa)
∂Q
× 10 3
∂∆P
Required
u ∆P (Pa)
10
20
30
40
1000
4271
8900
16023
3.06
2.16
1.4
1.2
± 2.5
± 10.7
± 22.3
± 40.0
Note that the required value of uncertainty in pressure drop is 0.25%.
COMMENT: The accuracy of pressure transducers is often reported as a percentage
of the full scale reading. To achieve the requirement of ± 2.5 Pa for a transducer
having a full scale reading of 20,000 Pa (3 psi) would require an uncertainty of ±
2.5/20,000 or 0.0125%. It is not likely that a single transducer could yield the desired
accuracy over the range of pressures required in the present problem.
PROBLEM 5.48
This problem is open-ended and developed by the user. Classroom/lab instructors will want
to use this problem as a basis for round-table discussions.
PROBLEM 5.49
This problem is open-ended and developed by the user. Classroom/lab instructors will want
to use this problem as a basis for round-table discussions.
PROBLEM 5.50
KNOWN:
=
σ FL
=
(t / 2) / I FL(t / 2) /( wt 3 /12)
L = 100 mm = 0.1 m
F = 980N
t = 0.01m
w = 0.03 m
FIND: maximum stress
SOLUTION
The best estimate for maximum stress is determined using the working values for each
variable. Plugging in the known values,
σ =
σ=
FL(t / 2) /( wt 3 /12) =×
196 106 Pa =
196 MPa
where I = wt 3 /12 =×
2.5 10−9 m 4 =
2500mm 4
All uncertainties can be treated as systematic errors. There are no statistics to compute
random errors (so Pσ =0). Then, the systematic errors in length are taken as ½ the ruler
increment:
BL = uL = 0.5 mm = 0.0005m
Bt = ut = 0.5 mm = 0.0005 m
Bw = uw = 0.5 mm = 0.0005 m
BF = uF = 10N
1/ 2
2
2
 ∂I
  ∂I  
BI =
Bw  +  Bt  

 ∂w   ∂t  
1/ 2
2
2
 t 3
  wt 2  
 B  + 
B  
=
 12 w   36 t  


5.89 × 10−11 m 4 =
58.9mm 4
=
1/ 2
2
2
2
2
 ∂σ
  ∂σ
  ∂σ   ∂σ
 
Bσ = 
BF  + 
BL  + 
Bt  + 
BI  
  ∂L
  ∂t
  ∂I
 
 ∂F
1/ 2
2
2
2
2
 Lt
  Ft
  FL   − FLt  
=  BF  +  BL  + 
Bt  + 
BI  
  2I
  2I   2I 2
 
 2 I
= 14, 739, 060 Pa
so,
1/ 2
 Bσ 2 + (tPσ ) 2 
uσ =


=
σ 196.0 ± 14.74 MPa
Bσ = 14.74 MPa
=
(95%)
or the uncertainty is about 7.5% of the maximum stress.
PROBLEM 5.51
 = f (T , T )
KNOWN: Q
1 2
T2 = 90o C
T1 = 180o C
B=
B=
0.2o C
T
T
1
2
S= S= 0.1o C
T1
T2
FIND: u  assuming (1) totally correlated and (2) totally uncorrelated errors
Q
ASSUMPTION: Large sample size
SOLUTION
Based on the measured mean values, the mean value of the result in heat transfer rate is
 = 5(T − T ) = 5(180 − 90C) = 450kJ / s
Q
1
2
Now, u  = f (u T , u T )
Q
1
2
1/ 2


where=
u   B  2 + (tP  ) 2  with t95 ~ 2 (value is consistent with large sample size and
Q
Q 
 Q
engineering practice – e.g., see PTC 19.1).
The standard random uncertainty is estimated by
1/ 2
2
2
 

  ∂Q
 
∂
Q
 
P  
P  +
P =
=
 ∂T T1   ∂T T2  
Q
1
  2
 

where PT = S and PT = S .
1
T1
2
1/ 2
2
2

 
 
0.71kJ / s
 5PT  +  −5PT =
 
2  
 1  

T2
For totally uncorrelated systematic errors, the systematic errors are independent of each
other:
1/ 2
2
2
1/ 2
 
2
2

  ∂Q
 

∂Q







B=
B  +
B =
1.41kJ / s
 5BT  +  −5BT =
 

 ∂T T1   ∂T T2  
Q
1
2  


uncor

1
2


 
 

For correlated errors, the errors will move in the same direction (high or low) and are
corrected from the uncorrelated estimate,
1/ 2
B
Qcor
2
2
 


  ∂Q
 
  ∂Q

 ∂Q
∂Q


B B 
B  +
B  + 2
= 
 ∂T T1   ∂T T2 
 ∂T  ∂T  T1 T2 
  2

 1  2 
 1

1/ 2
 5 × 0.2 2 + −5 × 0.2 2 + 2(5)(−5)(.2)(.2) 
=
(
) (
)


=0 kJ/s
So,
1/ 2


=
u   B  2 + (tP  ) 2 
Q
Q
Q


1/ 2


=  B  2 + (2P  ) 2 
Q
Q


u
= 2 kJ/s
such that
 450 ± 2kJ / s (95%)
=
Q
u
= 1.4 kJ/s
such that
 450 ± 1.4kJ / s (95%)
=
Q
Quncor
Qcor
COMMENT
The errors could be uncorrelated if (1) the different thermocouples came from different
batches or manufacturers, or (2) the thermocouples were calibrated against different
standards or methods. A portion of the uncorrelated error could come the use of different
measuring devices, which themselves were calibrated against different standards. If so, it is
possible to assign an uncorrelated portion and a correlated portion of the systematic error and
proceed as above.
On the other hand, if the thermocouples came from the same spool of wire (fairly common)
and/or were calibrated against the same standard, the errors are correlated.
PROBLEM 5.52
KNOWN:
R = p2/p1
p2 = 54.7 MPa
p1 = 42.0 MPa
0.50 MPa
B=
B=
p
p
2
FIND:
1
Compare results if errors are (1) uncorrelated and (2) correlated
SOLUTION
For values given:
R = p2 / p1 = 54.7/42.0 = 1.30
1/ 2
2
2
1/ 2

2
2
  ∂R
 

∂
R




BR
B  +
B   =  θ p Bp  +  θ p Bp  
= 
 ∂p p1   ∂p p2  
uncor
 1 1   2 2  
  2
 
 1
−p
1
where θp = 2 = 0.031 and θp = = 0.025, and
2
1
2
p1
p1
1/ 2
BR
cor
2
2


  ∂R

 ∂R  ∂R 
∂
R


B B 
= 
B  +
B  + 2
 ∂p p1   ∂p p2 
 ∂p  ∂p  p1 p2 
  2

 1  2 
 1

1/ 2
2
2


 

 

=  θp Bp  +  θp Bp  + 2  θp  θp  Bp Bp 
 1  2  1 2 
 1 1   2 2 
So,
R = 1.30 ± B (95%) where
1/ 2
if B = BR
uncor
2
2
= ( 0.031× 0.5MPa ) + ( 0.024 × 0.5MPa ) 


= 0.020 MPa
but
1/ 2
 0.031× 0.5MPa 2 + 0.024 × 0.5MPa 2 + 2(−0.031)(0.024)(0.5)(0.5) 
=
(
) (
)


cor
= 0.0035 MPa
if B = BR
The effect of having correlated errors nearly cancels out the effects of the individual
systematic errors.
The functional relationship can have an impact on whether correlated errors have an effect on
the uncertainty. Here R = p2/p1 allows for a decrease.
PROBLEM 5.53
KNOWN:
m = 22kg
V = 8 m/s
Bm = 0.001 kg
BV = 0.27 m/s
KE = 717 N-m
FIND:
SKE = 60.7 N-m with N = 24
uKE @ 68% and uKE @ 95%
SOLUTION
The confidence level in the uncertainty affects the value reported but not the fundamental
methods used to assess uncertainty. The “expanded” uncertainty (term used in PTC 19.1 –
2005) is the uncertainty at 95%,
1/ 2
 B2 + (t P) 2 
u=
KE
95


(95%)
while the combined standard uncertainty is the uncertainty at 68%.
1/ 2
=
u KE (B / 2) 2 + (P) 2 
(68%)
So find B and P:
Beginning with KE = ½ mV2,
1/ 2
2
2
 ∂KE
  ∂KE
 
=
BKE 
B  +
B  
 ∂m m   ∂V V  


1/ 2
= 0.00322 + 47.52 
(
) (
2

=  V 2 Bm / 2 + mVBV

= 47.5 N-m
P=
S= SKE / =
N 60.7N − m / 24 =12.4 N-m
KE
KE
with ν =23 . Then, t23,95 = 2.06 ~ 2
So, we can correctly state at either confidence level:
1/ 2
 47.52 + (2 ×12.4) 2 
u=
KE
= 53.5 N-m (95%)
1/ 2
=
u KE (47.5 / 2) 2 + (12.4) 2 
= 26.8 N-m (68%)
) 
1/ 2
2
PROBLEM 5.54
σmax = Ktσo ; σo = F/A ; A = (w – d)t
KNOWN:
FIND: Uncertainty in σmax
ASSUMPTION: Uncertainty in Kt is negligible
SOLUTION:
Develop an expression for σmax in terms of its known variables:
σ max = K t
uσ max = ±[(
∂σ max
∂w
∂σ max
∂d
∂σ max
∂t
∂σ max
∂F
=−
=
∂σ max
∂w
uw )2 + (
KtF
t(w − d)
KtF
t(w − d)
=−
=
F
(w − d)t
2
2
Kt
(w − d)t
∂d
ud ) 2 + (
∂σ max
∂t
ut ) 2 + (
∂σ max
∂F
u F ) 2 ]1 / 2
= 7.82x1010
= 7.82x1010
FK t
(w − d)t
∂σ max
2
= 117.3x10 9
= 58.7x10 3
where d = 0.5w ; Kt = 2.2 ; F = 10,000 N; w = 1.5 cm = 0.15 m ; t = 0.5 cm = 0.005 m
and uw = 0.02 cm = 0.0002 m ; uF = 500 N; ud = 0.5uw
u σmax = ±[(15.6x10 6 ) 2 + (17.6x10 6 ) 2 + (23.5x10 6 ) 2 + (29.3x10 6 ) 2 ] = 44.3x10 6 N/m 2
so putting it all together,
uσmax = 26.7 x 107 ± 44.3 x 106 N/m2
(95%)
PROBLEM 6.1
KNOWN:
A current loop having
N = 20
A = 1 in2 = 0.000645 m2
I = 0.02 A
B = 0.4 Wb/m2
FIND:
Torque on the current loop, Tµ.
ASSUMPTIONS:
The magnetic field is oriented at an angle of 90o to the current flow direction.
SOLUTION:
From (6.3)
Tµ = NIAB sin α
The maximum torque occurs when sin α = 1, which yields
Tµmax = NIAB
= 20 ( 0.02 A ) ( 0.000645 m 2 )( 0.4 Wb/m 2 )
max
Tµ=
1.03 × 10−4 N-m
PROBLEM 6.2
KNOWN:
FIND:
A voltage dividing circuit with RT = 5000 Ω (as shown in Figure 6.16)
Rm such that the loading error is less than 12% of the full scale value.
SOLUTION: Since the full-scale output is Ei (with e as the error)
e
=
Ei
R

R1 − RT + ( RT − R1 ) 1 R + 1
m
R2
 R

RT +  T R − RT  1 R + 1


m
1
A plot of the error as a function of meter resistance and R1 is shown below
Rm/RT = 1
Rm/RT = 2
Rm/RT = 4
PROBLEM 6.3
KNOWN: RT = R1 + R2 = 500 Ω =
Rm 10, 000 Ω
=
R1 kR
=
k 0.5=
Ei 10 V
T
FIND:
a) Loading error as a percentage of the output
b) Loading error as a percentage of the full scale output
SOLUTION:
The loading error may be defined as
e = E o′ − E o
and in terms of a percentage of the output
E o′ − E o E o′
=
−1
Eo
Eo
which yields, in terms of k
 (1 − k )  RT

+ 1 − 1
k 1 +
k
k  Rm


The loading error for this condition is 0.0125 or 1.25%.
As a percentage of full-scale output, Ei ,
E o′ − E o
=k−
Ei
k

 kR
k + ( 1 − k ) T + 1

 Rm
which produces a value for loading error of 0.62%. Since Eo is one-half of Ei, the loading errors are each 0.062
Volts.
PROBLEM 6.4
KNOWN:
R3 = R4 = 200 Ω
R2 = variable resistor
R1 = 40 x + 100 Ω
FIND:
a) R2 to balance the bridge with x = 0
b) Find a general expression for x = f ( R2 )
ASSUMPTIONS: Zero Galvanometer error
SOLUTION: At balanced conditions
R2 R4
=
R1 R3
Thus =
R2
(1)(100 )
and =
R2 100 Ω
and in general R2 = 40 x + 100 Ω .
PROBLEM 6.5
KNOWN:
A Wheatstone bridge with R1 = 20 x (x is a measured variable)
2
R=
R=
100 Ω R=
46 Ω at balanced conditions
3
4
2
FIND:
x
SOLUTION: Since at balanced conditions
R2 R4
=
=1
R1 R3
Thus
46
=1
20 x 2
x=
46
= 152
.
20
PROBLEM 6.6
KNOWN: A sensor has a resistance of 500 Ω under conditions of no load, and a static sensitivity of 0.5
Ω/N. The bridge circuit has
R=
R=
R=
R=
500 Ω
1
2
3
4
FIND:
(initially)
a) Eo for applied loads of 100, 200, and 350 N.
b) I1
c)
Repeat parts (a) and (b) with
=
Rm 10 kΩ
=
Rs 500 Ω
SOLUTION: For a bridge in which all resistances are initally equal
δR
Eo
R
=
E i 4 + 2 δR
R
(
)
and with a load of magnitude FL
δR = 0.5FL
0.5FL / 500
Eo
=
 0.5FL 
Ei

4 + 2
 500 
This yields
FL [N]
δR [Ω]
Eo [V]
100
50
0.238
200
100
0.455
350
150
0.652
The current flow through the sensor is I1 and for an infinite meter resistance is
I1 =
Ei
R1 + R2
which yields
R1+R2[ Ω ]
FL[N]
I1 [Amps]
100
1050
0.00952
200
1100
0.00909
350
1150
0.00870
c) Consider the circuit shown below, with values of δR equal to those for part (a)
im
R1
R2
i2
i1
Eo
i3
i4
10 kΩ
R3
is
600 Ω
Ei =10 V
A circuit analysis of this bridge yields the following simultaneous equations governing the currents and
potentials:
Ei = I s Rs + I1 ( R1 + R2 ) − I m R2
I1 R1 + I m=
Rm − I 3 R3 0
0
I m Rm + I=
4 R4 − I 2 R2
I m + I3 − I 4 = 0
I s = I1 + I 3
Ei = I s Rs + I 3 R3 + I 4 R4
Eo = I m Rm
Solving these equations simultaneously yields
R1 [Ω] Eo [V]
I1 [A]
550
0.104
0.0044
600
0.201
0.0042
650
0.291
0.0041
PROBLEM 6.7
KNOWN:
FIND:
Bridge circuit of Figure 6.36
Show that
( R)
C2 = C1 R1
2
is the requirement for a balanced bridge.
SOLUTION:
With impedances (for an AC circuit)
Z1 R1=
Z 2 R=
Z3
=
2
1
Z4
=
jω C1
1
jω C2
from
 Z3
Z1 
=
−
Eo Ei 

 Z 3 + Z 4 Z1 + Z 2 
For this case


1
1 

=
Eo Ei 
−

C
R
1 + 1 C 1 + 1 R 

2
2
which yields for Eo = 0
 R1 
C1 = C2  
 R2 
PROBLEM 6.8
KNOWN:
A bridge circuit is to be used to calibrate a 500 Hz frequency source. The bridge resonance
frequency is f =
FIND:
1
2π LC
Bridge components L and C to yield a resonance frequency of 500 Hz.
SOLUTION:
The product LC must be (1000π)2 and
500 =
1
2π LC
which yields
LC = 9.87 × 10 6 sec 2
Combinations of L and C which yield the appropriate resonance frequency are shown below.
1.00E+09
1.00E+08
Inductance (H)
1.00E+07
1.00E+06
1.00E+05
1.00E+04
1.00E+03
1.00E-02
1.00E+02
1.00E-01
1.00E+00
1.00E+01
Capacitance (F)
1.00E+02
1.00E+03
1.00E+04
PROBLEM 6.9
KNOWN: A Wheatstone bridge wth
FIND:
R1 = 200 Ω
R2 = 400 Ω
R3 = 500 Ω
R4 = 600 Ω
Ei = 5 V
a) Eo
b) Eo for R1 = 250 Ω.
ASSUMPTIONS:
The meter resistance Rg is infinite.
SOLUTION:
From (6.14)
 R1
R3 
500 
 200
−
−
=
− 0.606 V
5
Eo =
Ei 
=
 200 + 400 500 + 600 
 R1 + R2 R3 + R4 
or with R1 = 250 Ω
500 
 250
5 
Eo =
−
=
− 0 .35 V
 250 + 400 500 + 600 
COMMENT: Clearly the bridge output is non-linear with R1 over this range of values for R1.
PROBLEM 6.10
KNOWN:
A Wheatstone bridge having all resistances 500 Ω.
FIND:
Plot the output voltage for :
a) R1 varies from 500 to 1000 Ω
b) R1 and R2 change equally and in opposite directions between 500 and 600 Ω
c) R1 and R3 change equally over the range 500 to 600 Ω
SOLUTION:
The bridge output is given by
 R1
R3 
=
−
Eo Ei 

 R1 + R2 R3 + R4 
1.8
1.6
Output Voltage (V)
1.4
1.2
1
0.8
0.6
0.4
0.2
0
500
550
600
650
700
750
R1 (Ohms)
800
850
900
950
1000
0.5
0.4
0.3
0.2
0.1
0
-0.1
-0.2
-0.3
-0.4
-0.5
480
620
600
580
560
540
520
500
500
520
540
560
R1
c)
Identically Zero!
580
600
480
620
R2
Output (V)
b)
Eo
PROBLEM 6.11
KNOWN:
Potentiometer as shown in Figures 6.10 and 6.11.
Ei = 10 ± 0.1 V
RT = 100 ± 1 Ω
Rg =
100 Ω
Rx =
reading ± 2%
Null condition has negligible error.
FIND:
Design stage uncertainty in a measured value of voltage at nominal values of 2 and 8 V.
ASSUMPTIONS:
SOLUTION:
Loading errors should be included in the analysis.
From 6.35 (see Figure 6.16)
Eo
=
Ei
1

R R
1 + 2  1 + 1
R1  R g

Then at the design stage
2
 ∂ E
 
o
uEo =  ∑ 
u xi  
  ∂ xi
 
1
2
For this case Eo = f(Ei, Rx, Rg, RT) and RT = R1 + R2. At
Eo = 2 V R1 = 23.6 Ω
Eo = 8 V R1 = 88.3 Ω
R2 = 76.4 Ω
R2 = 11.7 Ω
The sensitivity indices may be evaluated as
∂E o
=
∂E i
at 2 V
∂ Eo
= 2,
∂ Ei
and at 8 V
∂ Eo
=8
∂ Ei
Ei

R R
1 + 2  1 + 1
R1  R g



− Ei R1  R1 + 1
R
∂Eo
∂Eo
∂Eo
g


at 2 V
0.021
at 8 V
0.137
=
=
=
2
∂R2  R  R
∂R2
∂R2


1 + 2  1 + 1 
 
 R1  Rg
 −R  R
 R  1
Ei  22  1 + 1 + 2 

 R R
1  g
∂Eo
 R1  Rg


= 2
∂R1
 R R

1 + 2  1 + 1 
 
 R1  Rg

 
 
at 2 V
∂Eo
∂Eo
=
−0.055
=
−0.00096
at 8 V
∂R1
∂R1
 − R2 
E

Rg2  i
∂Eo
∂Eo
∂Eo

=
=
−0.0031
=
−0.0075
at 2 V
at 8 V
2
∂Rg  R  R
∂Rg
∂Rg


1 + 2  1 + 1 
 R1  Rg
 
The uncertainty in R2 is estimated as
R=
RT − R1
2
1
uR2T + uR21  2
uR2 =
at 2 V uR2 =
1.11 Ω at 8 V uR2 =
2.03 Ω
At 2 V the uncertainty is estimated as
1
2
2
2
uEo= ( 2 × 0.1) + ( 0.021× 1.11) + ( −0.055 × 0.47 )  2


= ±0.203 V
And at 8 V as
1
2
2
2
uEo =( 8 × 0.1) + ( 0.137 × 2.03) + ( 0.00096 ×1.8 )  2


= ±0.847 V
PROBLEM 6.12
KNOWN:
Lissajous diagrams for the specific phase relations
See Figures below.
(a) In Phase
1.5
1
0.5
0
y
-1.5
-1
-0.5
0
0.5
1
1.5
-0.5
-1
-1.5
x
(b) 90 degrees out of phase
1.5
1
0.5
0
-1.5
-1
-0.5
0
0.5
1
1.5
-0.5
-1
-1.5
X
180 degrees out of phase
1.5
1
0.5
0
y
SOLUTION:
Y
FIND:
Sinusoidal inputs having specific phase relations
-1.5
-1
-0.5
0
-0.5
-1
-1.5
x
0.5
1
1.5
PROBLEM 6.13
KNOWN: sin φ =
FIND:
yi
ya
Show that the phase relationships can be determined from a Lissajous diagram using the equation
above.
SOLUTION:
Assume that the two sine waves with a phase delay of φ are the input to the x and y terminals of an oscilloscope.
Referring to Fig. 6.38, the figure below shows ya and yi.
When y = ya then the y signal is a maximum, with sin ω t = 1 and with y = A sin ω t and ya = A . At the yintercept where x=0 yi = A sin φ . Eliminating the amplitude A between these equations,
yi A sin φ
=
= sin φ
ya
A
1.5
1
ya
0.5
yi
y
0
-1.5
-1
-0.5
0
-0.5
-1
-1.5
x
0.5
1
1.5
PROBLEM 6.14
KNOWN:
FIND:
Phase lag occurs in electronic circuits
Design an arrangement which uses Lissajous diagrams to determine phase lag
SOLUTION:
The arrangement shown below will allow measurement of phase lag in an electronic circuit.
See Problem 6.15 for representative Lissajous diagrams.
Electronic Circuit
Reference
Source
Y input
X input
To Oscilloscope
PROBLEM 6.15
KNOWN:
Two sinusoidal signals are to be compared using a dual trace oscilloscope. The frequency ratios
are
a) 1:1 b) 1:2 c) 2:1 d) 1:3 e) 2:3
FIND:
f)5:2
Construct appropriate Lissajous diagrams
SOLUTION:
a)
Frequency Ratio 1:1
1.2
1
0.8
0.6
0.4
0.2
0
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
-0.2
-0.4
-0.6
-0.8
-1
-1.2
b)
Frequency Ratio 1:2
1.2
1
0.8
0.6
0.4
0.2
0
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
-0.2
-0.4
-0.6
-0.8
-1
-1.2
0.2
0.4
0.6
0.8
1
1.2
c)
d)
Frequency Ratio 2:1
Frequency Ratio 1:3
1.2
1.2
11
0.8
0.8
0.6
0.6
0.4
0.4
0.2
0.20
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
0.8
1
-0.2
0
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
-0.4
-0.2
-0.6
-0.4
-0.8
-0.6
-1
-0.8
-1.2
-1
-1.2
0.2
0.4
0.6
1.2
e)
f)
Frequency Ratio 2:3
Frequency Ratio 1:3
1.2
1.2
1
1
0.8
0.8
0.6
0.6
0.4
0.2
0.4
0
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0.2
0
0.2
0.4
0.6
0.8
1
1.2
-0.2
0
-1.2
-1
-0.8
-0.6
-0.4
-0.2
-0.4
-0.2
-0.6
-0.4
-0.8
-0.6
-1
-1.2
-0.8
-1
-1.2
0
0.2
0.4
0.6
0.8
1
1.2
PROBLEM 6.16
KNOWN:
FIND:
RC filter:
=
k 1;=
f c 100 Hz
Attenuation at 10, 50, 75, and 200 Hz
ASSUMPTIONS:
Filter is of the low-pass, Butterworth type
SOLUTION:
For a low-pass RC Butterworth filter,
2k
M (=
f ) 1 1 + ( f f c ) 


0.5
Recall that “attenuation” means reduction. In the context of a filter, a device that reduces the output amplitude
of targeted frequencies M
( f ) < 1 it must correspond to a negative value of the dynamic error.
error is given by
Dynamic
δ=
( f ) M ( f ) −1
Recall also that a “gain” refers to a positive value of dynamic error.
For a low-pass RC Butterworth filter, M
( f ) < 1 always which is consistent with first-order system
behavior.
f (Hz)
M(f)
δ(f)
Atttenuation %
10
50
75
200
0.995
0.894
0.800
0.447
-0.005
-0.105
-0.200
-0.553
0.5
10.5
20.0
55.3
PROBLEM 6.18
KNOWN:
FIND:
LC low-pass Butterworth filter
-3 dB ≤ M(0 ≤ f ≤ 5 kHz)
M(f ≥ 10 kHz) < -30 dB
Values for L, C and k
SOLUTION
This problem is an open-ended design and one possible solution follows. For a low-pass,
Butterworth filter such as shown in Figure 6.31, we begin by fixing the cut-off frequency and
then estimating the order (number of reactive stages) k. We end by specifying the C and L
values. For the filter:
M(f) = 1/[1 + (f/fc)2k]1/2
We set fc = 5 kHz, such that M(5 kHz) = 0.707 = -3 dB, and meet one constraint of the
design.
The next step in the design is to determine the number of filter stages required to meet the
attenuation constraint at 10 kHz. For at least 30 dB attenuation
-30 dB = 20 log M(10 kHz)
or want
M(10 kHz) ≤ 0.0316
then,
M(10 kHz) ≤ 0.0316 = 1/[1 + (10 kHz/5 kHz)2k]1/2
This gives k = 4.98 ≈ 5. Note: k can be found by trial and error or by direct estimate as
follows.
By trial and error:
k
M(10 kHz)
1
0.45
3
0.12
5
0.0312
By direct estimate, with f/fc = 10000/5000 = 2 and M = 0.0316:
1− M2
log(
)
2
M
k=
= 4.98
2log(f/f c )
A five-order filter has five reactive elements. To specify the capacitors and inductors, we
refer to Table 6.1 with its scaling equations (6.60)
1
C = Ci
2πf c R
L = Li
R
2πf c
Hence, for a normalized value of R = 1Ω and fc = 5,000 Hz values for C and L yield:
C1 =
20µf
C3 =
64µf
L2 =
52 µh
L4 =
52 µh
C5 = 20µf
PROBLEM 6.19
KNOWN:
Circuit of Figure 6.39.
FIND:
Thevenin equivalent.
In open-circuit operation,
Eo = InRn
In short-circuit operation,
In = Eth/Rth
or,
Eth = InRn
and
Rth = Eth/In
PROBLEM 6.20
KNOWN:
Z1 = 500Ω
FIND:
E1/Em
Zm = 100,000Ω
SOLUTION
From Figure 6.17 and corresponding equation (6.38)
Em/E1 = 1/(1 + Z1/Zm) = 1/(1 + 500/100,000) = 0.995
So,
E1/Em = 1.005
eI = 0.005E1
Loading error is about 0.5% of E1.
PROBLEM 6.21
FIND:
Em/E1 versus Z1/Zm
SOLUTION
From Figure 6.17 and corresponding equation (6.38)
Em/E1 = 1/(1 + Z1/Zm)
The results for a range of independent values of Z1 /Zm is plotted below.
1.0
Em/E1
0.8
0.6
0.4
0.2
0.0
0.000001
0.0001
0.01
1
Z1/Zm
As the loading error goes to 0 as Em/E1 goes to 1, Z1 should be much less than Zm to
minimize loading error. To reduce error to 1% (which is still a significant error), the ratio
must be at least 1:100; to 0.01% (a good goal), the ratio must be at least 1:10,000 and so
forth.
PROBLEM 6.22
KNOWN: Z1 =
1×109 Ω
FIND: Z m required to keep EI E ≤ 0.1%
SOLUTION:
For the circuit in Figure 6.40, Equation 6.38 can be written
=
Em E1Z m
( Z m + Z1 )
where E1 is the Thevenin equivalent voltage. For =
Z1 10 Ω
9
=
Em E1 Z m
(Z
m
+ 109 )
Then the relative loading error or signal attenuation is
eI E1 = 1 − Em E1
Zm Ω
6
10
109
1012
(6.39)
(1 − Em
E1 ) %
99.9%
50%
0.1%
Z m must be at least 1012 Ω to meet the constraint. This requires=
Z m Z1 1012 /109 ≥ 1000
COMMENT: Loading error is only a problem if it goes undetected. It can be prevented by proper design
as it is easily predicted.
PROBLEM 6.23
KNOWN:
Circuit of Figure 6.41
FIND: Eo Ei for the circuit
SOLUTION:
Across the divider, (6.8) gives
0.375 Ei or
=
Eo E=
i ( 60 ) 160
( E=
o Ei ) divider
0.375
This is equivalent to a gain of
G [ dB] = 20 log ( Eo Ei ) = −8.5 dB
The overall circuit gain based on Gamp =
+32 dB G filter =
−12 dB , and the divider circuit is
Eo Ei =
( 32 − 12 − 8.5)
dB = 11.5 dB = 3.76
PROBLEM 6.24
KNOWN: Temperature measurement system employing a resistance temperature detector
and a Wheatstone bridge circuit having the following characteristics
R =Ro 1 + α ( T − To ) 
Ro =
100 Ω at 0C R3 =R4 =500 Ω
There are two error sources, the sensitivity of the galvanometer, and the accuracy of the
fixed resistors.
FIND: Design a combination of uncertainty in the fixed resistors and the current flow
through the galvanometer to provide an uncertainty in temperature of ±1°C
SOLUTION:
The sensitivity indices are needed to determine the required accuracy of the detector
resistance measurement to yield an uncertainty of ±1°C, and the uncertainty of the
measurement of R1 in the bridge. The required level of uncertainty in the resistance
measurement can be determined from
∂R
∂
 Ro 1 + α ( T − To =
=
)   α Ro
∂T ∂T  
Knowing that uT = ±1 C the required uncertainty in R1 is found from
∂R
=
uR
=
uT α=
RouT ( 0.00395 )(100=
)(1) 0.395 Ω
∂T
Then to evaluate the sensitivity of R1 to the uncertainty in Ig sequential perturbation can be
employed with equation 6.22 to yield a value of 0.000045 A/Ω or 22.2 Ω/mA. The
uncertainty in R1 may be expressed
uR1 =
where
 ∂R1
uI g

 ∂I g
2
  ∂R1
  ∂R
  ∂R

uR2  +  1 uR3  +  1 uR4 
 + 
  ∂R3

  ∂R4
  ∂R2
2
2
2
∂R1 R3
∂R1 R2 ∂R1 − R2 R3
=
=
=
∂R2 R4
∂R3 R4 ∂R4
R42
With appropriate assumptions of uncertainties in the values of the R’s and Ig the value of uT
can be evaluated, and an iterative process used to satisfy the constraint.
PROBLEM 6.25
FIND:
Design low pass filter with fc = 100 Hz
Want M(50 Hz) ≥ 0.95; M(200 Hz) ≤ 0.01
KNOWN:
RL = RS = 10 Ω
SOLUTION
For a Butterworth filter having k reactive elements, the frequency response is given by
M(f) = 1/[1 + (f/fc)2k]1/2
rearranging and solving for k:
1− M2
log(
)
2
M
k=
2log(f/f c )
In terms of f/fc, where fc = 100 Hz, we want M(50/100 = 0.5) ≥ 0.95; M(200/100 =2 ) ≤ 0.01.
The highest value of k will satisfy both conditions. Solving yields k(f/fc = 0.5) ≥ 2 and k(f/fc
= 2) ≥ 4. Choose k ≥ 4.
If we select a k = 5 order filter, there will be 5 reactive elements. To specify the capacitors
and inductors, we refer to Table 6.1 with its scaling equations (6.60)
1
C = Ci
2πf c R
L = Li
R
2πf c
Hence, for a normalized value of R = Rs/1Ω = RL/1Ω = 10Ω and fc = 100 Hz values for C and
L yield:
C1 =
98µf
C3 =
318 µf
L2 =
26 mh
L4 =
26 mh
C5 = 98µf
PROBLEM 6.26
FIND:
Design high pass filter with fc = 5000 Hz
Want M(2500 Hz) ≥ -20dB; M(8000 Hz) ≤ -0.45 dB
KNOWN:
RL = RS = 10 Ω
SOLUTION
For a high pass Butterworth filter having k reactive elements, the frequency response is given
by equation (6.61):
M(f) = 1/[1 + (fc/f)2k]1/2
rearranging and solving for k:
1− M2
log(
)
M2
k=
2log(f c /f )
Since dB = 20 log M(f), then: M(8000) ≥ 0.95 and M(2500) ≤ 0.1.
In terms of f/fc, where fc = 5000 Hz, we want M(8000/5000 = 1.6) ≥ 0.95; M(2500/5000 =
0.5 ) ≤ 0.1. The highest value of k will satisfy both conditions. Solving yields k(f/fc = 0.5) ≥
3.3 and k(f/fc = 1.6) ≥ 2.36. Choose k ≥ 4.
If we select a k = 5 order high pass filter, there will be 5 reactive elements. To specify the
capacitors and inductors, we refer to Table 6.1 with its scaling equations (6.60)
1
C = Ci
2πf c R
L = Li
R
2πf c
Hence, for a normalized value of R = Rs/1Ω = RL/1Ω = 10Ω and fc = 5000 Hz values for C
and L yield:
C1 =
2µf
C3 =
6.4 µf
L2 =
515 mh
L4 =
515 mh
C5 = 2µf
PROBLEM 6.27
FIND:
Design low pass filter with fc = 10,000 Hz using 741 op-amp
KNOWN:
C2 = 0.1µF
K = 20
first-order (only one reactive element)
SOLUTION
This design problem has an open-ended solution. One possible solution follows.
Use the design of Figure 6.33a.
The cut-off frequency is fc =
1
2π R2 C2
= 10,000 Hz
1
= 159 Ω
2π f c C2
The gain is K = R2/R1 = 20, so R1 = R2/20 = 8 Ω
With C2 = 0.1µF = 100 nF, R2 =
Program Low Pass Butterworth Active Filter can also be used to complete this problem. An
advantage with the software is that it provides a ready visual on the effects of changing the R
and C values on the gain and cut-off frequency. Output for this problem is shown.
PROBLEM 6.28
FIND:
Design high pass filter with fc = 10,000 Hz using 741 op-amp
KNOWN:
C1 = 0.1µF
K = 10
first-order (only one reactive element)
SOLUTION
Use the design of Figure 6.33b.
The cut-off frequency is
fc =
1
= 10,000 Hz
2π R1 C1
With C1 = 0.1µF,
R1 =
1
= 160 Ω
2π f cC1
The gain is K = R2/R1 = 10, so
R1 = R2/10 = 16 Ω
PROBLEM 6.29
Program Oscilloscope offers the user a very simple but functionally correct two-channel
oscilloscope. Two input signals are provided. The user can vary the time sweep and the
amplifier gains. A trigger is available off of Channel B and the effect of changing trigger
levels can be explored.
PROBLEM 6.30
Program Butterworth Filters Overall can be used to examine the effect of using a low pass, a
high pass or a bandpass Butterworth filter on a signal of variable frequency. The output
signal and output amplitude spectra are displayed.
For example, with a signal frequency of 400 Hz, the lowpass (fc = 300 Hz) and highpass (fc =
500 Hz) filters show a definite attenuation (~ 0.8V amplitude for lowpass and ~ 1V for high
pass) whereas the bandpass filter (300 < fc < 500 Hz), passes the full signal amplitude (~ 2V
or 0dB).
PROBLEM 6.31
Program Bessel Filters Overall can be used to examine the effect of using a low pass, a high
pass or a bandpass Bessel filter on a signal of variable frequency. The output signal and
output amplitude spectra are displayed.
For example, with a signal frequency of 400 Hz, the lowpass (fc = 300 Hz) and highpass (fc =
500 Hz) filters show a definite attenuation (~ 0.6V amplitude or -6 dB) whereas the bandpass
filter (300 < fc < 500 Hz), passes the full signal amplitude (~ 2V or 0dB).
PROBLEM 6.32
Program LP Butterworth Noise can be used to study the effect of using a low pass
Butterworth filter on a sinusoidal signal of variable frequency and filter order number. A
high frequency noise signal is superimposed on the sine wave. The output signal and output
amplitude spectra are displayed.
The signal becomes cleaner as the number of stages is increased. k= 4 or 5 work well.
Observe how the signal is shifted in time as k is increased (watch the value near t = 0 shift).
Increasing the stages, increases the phase shift. This may or may not be important in an
application.
PROBLEM 6.33
Program Monostable simulates an operating monostable integrated circuit. It is called a oneshot for the reason seen in its output signal: when activated, it produces a single square
waveform of time duration t.
The three output charts are:
- trigger value: the trigger is HIGH until engaged (LOW). Internally, transistor T2 is on
until the trigger is engaged whereby T2 goes off momentarily allowing a jump in the
output voltage.
- capacitor output value: on trigger, the capacitor fires. Note the width of the output
signal is coincident with the capacitor value.
- circuit output signal: the monostable provides a single square wave having a duration
determined by the R and C values.
Different values will create a 1 s duration wave. For example, 500Ωand 1.82µF work.
PROBLEM 6.34
Program 741 Opamp simulates an inverting operating amplifier integrated circuit. The user
can vary the gain characteristics.
The opamp is inverting. A negative value of Vin produces a positive value of Vout.
The opamp has a gain given by K = R2/R1. Any combination of R2 = 5R1 will yield a gain of
5. Likewise, any combination of R2 = 0.5R1 yields a gain of 0.5 (an amplifier is called an
attenuator when 0 < G < 1).
The opamp’s output is governed by its operating voltage, +/- 15V here. This unit shows a +/14V maximum output.
PROBLEM 7.1
KNOWN:
E(t) = 5sin 2πt mV
FIND:
Convert to a discrete time series and plot
SOLUTION
The signal is converted to a discrete time series for using N = 8 and sample time
increments of 0.125, 0.30, and 0.75 s and plotted below. The time increments of 0.125 and
0.30 s produce discrete series with a period of 1s or frequency of 1 Hz. The series created
with a time increment of 0.75 s, which fails the Sampling Theorem criterion, portrays a
signal with a different frequency content; this frequency is the alias frequency.
6
DISCRETE SERIES
4
dt = 0.125 s
2
0
0
0.2
0.4
0.6
-2
-4
-6
TIME [s]
0.8
1
1.2
6
dt = 0.30 s
DISCRETE SERIES
4
2
0
0
0.5
1
1.5
2
2.5
3
-2
-4
-6
TIME [s]
DISCRETE SERIES
6
dt =0.75 s
4
2
0
0
1
2
3
-2
-4
-6
TIME [s]
4
5
6
PROBLEM 7.2
KNOWN:
FIND:
Repeat Problem 7.1 using N = 128 points.
The discrete Fourier transform for each series.
SOLUTION
The DFT for the time series representations of E(t) using N = 128 and δ t
= 0.125, 0.30, and 0.75 s, respectively was executed using a DFT algorithm
(any such algorithm on the companion software or using the approach described in Chapter 2
will work) and shown below. The DFT returns an exact Fourier transform of the discrete time
series but not necessarily the time signal from which it is based. Whether this DFT exactly
represents E(t) depends on the criteria:
(1)
(2)
fs = 1/ δ t > 2f
m/f1 = N δ t
m = 1,2,3, ...
With f = 1 Hz and f1 = f: (a)
δ t = 0.125 s and N = 128:
fs = 1/0.125s = 8 Hz > 2 Hz
m/1 Hz = 128/0.125 or m = 16 an exact integer value.
Another way to look at this second criterion: the DFT resolution δ f = 1/N δ t =
0.0625 Hz, to which 1 Hz is an exact multiple.
Both criteria are met. Therefore, this DFT will exactly represent E(t) in
both frequency and amplitude, as shown below.
(b)
δ t = 0.3 s and N = 128
fs = 1/0.3s = 3.3 Hz > 2 Hz
m/1 Hz = 128/0.3 or m = 38.4 not an exact integer value.
Criterion (1) is met but Criterion (2) is not met. Therefore, an alias frequency will not
appear. But this DFT will not exactly represent E(t) in amplitude and spectral leakage will
occur, as seen below. So we find an amplitude less than 5 at a frequency centered at 1 Hz.
δ t = 0.75 s and N = 128
(c)
fs = 1/0.75s = 1.33 Hz < 2 Hz
m/1 Hz = 128/0.75 or m = 96 an exact integer value.
Criterion (1) is not met but Criterion (2) is met. Therefore, this DFT will not exactly
represent E(t) in frequency. However, it does represent the signal amplitude exactly. An alias
frequency at 0.33Hz appears in the DFT.
Note how the frequency resolution scale changes between plots.
6.0
dt=0.125
AMPLITUDE
5.0
4.0
3.0
2.0
1.0
0.0
0
1
2
3
4
5
AMPLITUDE
FREQUENCY [Hz]
4
3.5
3
2.5
2
1.5
1
0.5
0
dt=0.3
0
0.5
1
FREQUENCY [Hz]
1.5
2
6
dt=0.3
AMPLITUDE
5
4
3
2
1
0
0
0.2
0.4
FREQUENCY [Hz]
0.6
0.8
PROBLEM 7.3
KNOWN: T(t) = 2 sin 4πt oC
fs = 4 and 8 Hz; N = 128
FIND:
Compute the Fourier transform from the resulting series.
SOLUTION
The fundamental frequency of this signal is f1 = 2 Hz. From the
Sampling theorem, an appropriate sample rate is
fs > 2f1
or
fs > 4Hz
At fs = 4 Hz the Sampling Theorem is not upheld whereas at fs = 8 Hz the Sampling theorem
is upheld. For fs = 4 Hz, see the COMMENT below.
For amplitude fidelity, mT1 = N δ t which can be written as m/f1 = N/fs m = 1,2, ...
With fs = 8 Hz and N = 128, m = Nf1/fs = N δ t /T1 = 8 an exact integer. We can expect that
the resulting DFT will be an exact representation of T(t), as seen below.
3
AMPLITUDE
dt=0.125 N = 128
2
2
1
1
0
0
1
2
3
4
5
FREQUENCY [Hz]
COMMENT
The result of sampling at fs = 2f is somewhat interesting. A non-unique
reconstruction of T(t) can result which will depend wholly on the initial condition of T(t) at
the time sampling commences. Because of this, we can not write the Sampling Theorem
criterion as fs ≥ 2f as is sometimes found in texts. Try this.
PROBLEM 7.4
KNOWN:
Signal frequency, f1, and the sample rate, fs
FIND:
Any alias frequency, fa arising
SOLUTION
The Nyquist or folding frequency is defined as fN = fs/2. All frequencies in the sampled
signal that are above fN will be folded back to lower frequencies below fN, a process depicted
by the folding diagram (Figure 7.3).
(a) f1 = 60 Hz
fs = 90 Hz
fN = fs/2 = 45 Hz.
The ratio, f/fN = 1.33, that is f = 1.33fN. Referring to The folding diagram, a frequency of
1.33fN will be folded back to a frequency of 0.67fN. So f1 = 60 Hz but in the sampled signal it
behaves as fa = 0.67fN = 30 Hz and out-of-phase with the original signal.
(b) f1 = 1200 Hz
fs = 2000 Hz
fN = fs/2 = 1000 Hz.
The ratio, f/fN = 1.2, that is f = 1.2fN. Referring to The folding diagram, a frequency of 1.2fN
will be folded back to a frequency of 0.8fN. So f1 = 1200 Hz but in the sampled signal it
behaves as fa = 0.8fN = 960 Hz and out-of-phase with the original signal.
(c) f1 = 10 Hz
fs = 6 Hz
fN = fs/2 = 3 Hz.
The ratio, f/fN = 3.3, that is f = 3.3fN. Referring to The folding diagram, a frequency of 3.3fN
will be folded back to a frequency of 0.7fN. So f1 = 10 Hz but in the sampled signal it
behaves as fa = 0.7fN = 7 Hz and in-phase with the original signal.
(d) f1 = 16 Hz
fs = 8 Hz
fN = fs/2 = 4 Hz.
The ratio, f/fN = 4, that is f = 4fN. Referring to The folding diagram (and projecting its
behavior beyond the values shown), a frequency of 4fN will be folded back to a frequency of
0. So f1 = 16 Hz but in the sampled signal it behaves as fa = 0 Hz. It will be seen as a constant
(dc) signal.
PROBLEM 7.5
KNOWN:
FIND:
E(t) = sin 2πt + 2 sin 8πt
fs = 16 Hz
DFT (amplitude spectrum) of a discrete time series representation of E(t).
Reconstruct E(t) from the Fourier transform.
SOLUTION
Begin by generating a discrete time series. We can rewrite E(t) as
E(t) = sin 2πf1t + 2 sin 2πf2t
with f1 = 1 Hz and f2 = 4f1 = 4 Hz. An exact DFT representation of the E(t) will
result if the discrete time series is created using
(1) fs > 2f2
(2) mT1 = N δ t or m/f1 = N/fs
m = 1,2,...
For (1) to be true, fs > 8 Hz. This constraint is met, so frequency will be correctly
represented. For (2) to be true, we need
N = mT1/ δ t = mfs/f1 = m(16 Hz)/1 Hz = 16m
That is for amplitude to be correctly represented, N must be an exact multiple of 16 (e.g. 16,
32, ...). We use N = 32 to construct the discrete time series. The time series and its DFT are
computed using a DFT algorithm. The time signal is then reconstructed from the spectral
information by
=
E (t )
N/2
∑C
k =1
k
sin(2π f k t + tan −1
Ak
Bk
)
The DFT and the respective reconstructed time signal are shown below. The reconstructed
signal is exact as expected.
Try this problem at different sample rates.
2.5
dt=0.0625 N = 32
AMPLITUDE
2.0
1.5
1.0
0.5
0.0
0
1
2
3
4
5
6
7
FREQUENCY [Hz]
4
reconstruct
3
original
SIGNAL
2
1
0
-1
-2
-3
-4
0
0.2
0.4
0.6
TIME [s}
0.8
1
1.2
PROBLEM 7.6
KNOWN:
FIND:
y (t ) =
∑ (4 /(2n − 1)π ) sin[2π (2n − 1)t /10]
Appropriate values for fs and N δ t with f m ≤ 2 Hz and N = 2M.
SOLUTION
Filtering the signal at and below 2 Hz limits this series representation of y(t) to 5 terms. Each
nth term of the series has the frequency fn = (2n-1)/10. The fundamental frequency (n=1) is f1
= 0.1 Hz.
In order to properly construct a discrete time series, the following criteria should be met:
(1)
fs > 2fm
Sampling Theorem criterion for frequency content
(2)
mT1 = N δ t m = 1,2,...
Criterion for amplitude content
where T1 = 1/f1. Because fm = 2 Hz, criterion (1) is met by setting fs > 4 Hz.
There are many different correct solutions to this problem. One such solution is to set fs = 8
Hz. Then this requires,
m(10 s) = N(1/8 Hz)
m = 1,2,...
Also, we must set N = 2M where M is an integer. Criterion (2) is met with N
= 210 = 1024. Hence,
fs = 8 Hz
N = 1024
and
N δ t = N/fs = 1024/8 = 128 s
COMMENT
We stress that there can be different combinations of fs and N δ t that will meet the criteria of
(1) and (2) in a given problem. In a practical problem, additional restrictions, such as the
maximum sample rate available, the maximum data set size that can be stored, or the
available range of anti-alias filters will limit the number of possible combinations.
PROBLEM 7.7
KNOWN:
FIND:
Straight Binary Number
Equivalent base 10 representation
SOLUTION
(a) 1010 = 1*23 + 0*22 + 1*21 + 0*20 = 1010
(b) 11111 = 1*24 + 1*23 + 1*22 + 1*21 + 1*20 = 3110
(c) 10111011 = 1*27 + 0*26 + 1*25 + 1*24 + 1*23 + 0*22
+ 1*21 + 1*20 = 18710
(d) 1100001 = 1*26 + 1*25 + 0*24 + 0*23 + 0*22 + 0*21 + 1*20 = 9710
PROBLEM 7.8
SOLUTION
(a) 1100111.1101 into decimal
1100111 = 1*26 + 1*25 + 0*24 + 0*23 + 1*22 + 1*21 + 1*20 = +10310
.1101 = 1*2-1 + 1*2-2 + 0*2-3 + 1*2-4 = 0.812510
So 1100111.1101 = 103.812510
(b) 4B2F into binary
4B2F = 0100 1011 0010 1111 = 0100101100101111
(c) 278.63210 into binary
278 = 100010110
.632 = .101000011
So 278.63210 = 100010110.101000011
PROBLEM 7.9
FIND:
Equivalent two's complement representation
SOLUTION
For example, assuming a 12 bit binary number, two's complement assigns a 0 to the MSB is
(+) with succesive bits in straight binary (0 HIGH, 1 LOW), or a 1 if (-) with successive bits
complementary (1 HIGH, 0 LOW). Because two's complement has only a single zero add 1
to its count (negative number).
(a) 1010 = 0000 0000 1010
(b) -1010 = 1111 1111 0110
(c) -24710 = 1111 0000 1000
(d) 101310 = 0011 1111 0101
PROBLEM 7.10
KNOWN: Two's complement code; M = 8
SOLUTION
Two's complement code uses the MSB as the sign bit, with a 0 for a
positive number (+), and then straight binary. The largest positive number is:
0111 1111 = (+) 26 + 25 + 24 + 23 + 22 + 21 + 20
= (+) 64 + 32 + 16 + 8 + 4 + 2 + 1 = 12710
Adding a one to this binary number yields a negative number. Negative
numbers use complementary binary (0 = HIGH, 1 = LOW). Note: Because
two's complement has only a single zero add 1 to a negative count.
1000 0000 = (-)27 = (-)128 = -12810
PROBLEM 7.11
KNOWN: Two's complement code with M = 8
SOLUTION
Two's complement code uses the MSB as the sign bit, with a 1 for a
negative number, and complementary binary (0 = HIGH, 1 = LOW). Note:
Because two's complement has only a single zero, you add 1 to its negative count. The
largest negative number is:
1000 0000 = (-1) + 26 + 25 + 24 + 23 + 22 + 21 + 20
= (-1) + 64 + 32 + 16 + 8 + 4 + 2 + 1 = -12810
Subtracting one from this (the binary not the decimal) number:
0111 1111 = (+) 26 + 25 + 24 + 23 + 22 + 21 + 20
= (+1)127= +12710
PROBLEM 7.12
KNOWN: Dual-slope integration A/D
FIND: List error sources. Derive a relationship between the uncertainty in the
digital result and the slope of the integration process.
SOLUTION
Possible error sources include (see Figure 7.11): 1. errors in controlling t1 ; 2. errors in
measuring tm ; 3. errors in applied Eref
Since the slope is proportional to Eref, then Ei = Eref(tm/t1). The uncertainty in the measured
voltage can be expressed as:
 ∂E
i
± 
uE =
u
i
 ∂Eref ref


2
  ∂E
i
 +
ut

m

  ∂ (tm )
2
  ∂Ei
ut
 + 
i
  ∂ (ti )



2
1/ 2




The ratio tm/t1 could be treated as a single entity but we separate the two variables here.
The value of Eref, which determines the slope, then contributes to the uncertainty in two
ways: 1. The uncertainty in Eref will vary linearly with tm. 2. Because the sensitivity index for
the uncertainty in tm/t1 equals Eref, the slope directly affects the sensitivity of the measured
voltage through the uncertainty in time determination.
The devices that contribute to these uncertainties are associated with the frequency source
and the counter.
PROBLEM 7.13
KNOWN:
FIND:
A/D Converter
M = 4, 8, 12, 16
-5 to 5 V bipolar
Q, dynamic range
SOLUTION
The resolution of an M-bit A/D converter is expressed as
Q = EFSR/2M
For this problem, EFSR= 10 V (i.e. -5 to 5 V)
The dynamic range can be expressed in terms of the signal-to-noise ratio (SNR),
SNR [dB] = 20 log 2M
M
Q [V]
4
8
12
16
0.62500
0.03906
0.00244
0.00015
SNR [dB]
-24
-48
-72
-96
PROBLEM 7.14
KNOWN: A/D Converter: M = 12; EFSR = 5V; e1/EFSR = 0.0003
FIND: eQ, emax, uE/E
SOLUTION
Q = 5V/212= 1.2 mV
eQ= ± Q/2 = ± 0.6 mV
The maximum error that could be present in any measurement is found by
summing all known errors:
umax = eQ + (EFSR)(e1/EFSR) = 0.6 + 1.5 = 2.1 mV
However, the probable error provides a reasonable estimate:
uE= ± (eQ2 + e12)0.5 = ± (.62 + 1.22)1/2= 1.6 mV
(95%)
or uE/E = 0.032%.
COMMENT
A maximum error assumes that all possible errors are at their maximum value during a
measurement. It is different from a probable error.
PROBLEM 7.15
KNOWN: Single ramp A/D converter: M = 8; EFSR = 10 V; fclock = 2.5 MHz
Comparator: Eth = 1 mV
(i) Ei = 6.000 V (ii) Ei = 6.035 V.
FIND: Resolution. Binary output. Conversion times.
SOLUTION
(i) Resolution: Q = 10 V/28 = 39 mV (or 0.4% FSO) for a ramp slope equivalent to
0.039 V/time step.
For Ei = 6.000 V, the number of steps required is:
Ei/Q = 153.6 ≈ 154 steps
With a single count for each step, 15410 = 1001 1010 in straight binary.
(ii) Repeating for Ei = 6.035 V: Q = 39 mV.
Ei/Q = 155 steps or 15510 = 1001 1011 in straight binary.
For either case:
With fclock = 2.5 MHz, each step requires 0.4 µ s . So the conversion times are:
Ei = 6.000 V: t = (0.4 µ s )(154) = 61.1 µ s
The maximum conversion time required for a given input occurs at its maximum count or 28:
tmax = (0.4 µ s )(256) = 102.4 µ s .
The average conversion time is:
tavr = (0.4 µ s )(128) = 51.2 µ s
PROBLEM 7.16
KNOWN: A/D converter: M = 10; EFSR = 10 V. fclock= 1 MHz.
FIND: Compare conversion times for successive approximation versus dualramp operation.
SOLUTION
Successive approximation Method: A maximum conversion time would result after M trials.
For M = 10,
tmax = M/fclock= 10/1 MHz = 10 µ s
Dual-ramp methods: A maximum conversion time results after (2)(2M) = (2M+1) trials.
For M = 10,
tmax = 2M+1/fclock = 211/1 MHz = 2048 µ s .
PROBLEM 7.17
KNOWN: D/A converter: M = 8
Eo = 3.58 V if register = 10110011
FIND: Eo if register = 01100100.
ASSUMPTION: Straight binary code.
SOLUTION
10110011 = 17910 or 179 counts. For a known Eo this is equivalent to:
Q = 3.58 V/179 = 0.020 V
Then, for 01100100 = 10010 or 100 counts:
Eo = (0.020 V)(100) = 2.000 V.
PROBLEM 7.18
KNOWN: Successive approximation converter: M = 4; EFSR = 10 V;
Ei = 4.9 V
FIND: E*. Required M for Q ≤ 2.5 mV.
SOLUTION
(i) Resolution:
Q = 10 V/24 = 0.625 V So with Ei = 4.9 V, the number of register counts
is 4.9 V/0.625 V = 7.8. Referring to Figure 7.7, if we assume the A/D uses a straight
encoding scheme whereby eQ = 1 bit = 0.625 V, the register will show 7.8 ⇒ 7 counts, or 710
= 0111. Hence, E* = (7)(.625 V) = 4.375 V. Note that Ei- E*< eQ.
On the other hand: If the A/D used a bias shift (offset) encoding whereby eQ = ± 1/2 bit = ±
0.3125 V, the register will show 7.8 ⇒ 8 counts, or 810 = 1000. Hence, E* = 5.000 V. Again,
Ei – E* < eQ.
(ii) For |Ei – E*| = eQ < 2.5 mV: requires either M = 12 (for a straight encoding scheme) or M
= 11 (for a bias shifted scheme).
PROBLEM 7.19
SOLUTION
Resolution: Q = EFSR/2M regardless of the conversion method.
Maximum Conversion time: For a given clock speed, fclock,
Successive Approximation: t = M/fclcok
Ramp: t = 2M/fclock
Parallel: t = 1/fclock
For any converter, resolution improves with bit number equally. The trade-off is
seen in the required conversion times. Ramp times increase exponentially with
bit number. Parallel method conversion times are essentially independent of bit
number, but their component costs increase exponentially with bit number.
PROBLEM 7.20
KNOWN: A/D converter: M = 10; EFSR = 10 V; register = 1010 1101 11
FIND: Ei
SOLUTION
For a straight binary code, 1010 1101 11 = 69510 which is equivalent to
E*= (695)Q = (695)(10/210) = 6.787 V
Assuming that eQ= Q = 1 LSB,
Ei = E* + 1 LSB = 6.787 + 1 LSB or 6.787 ≤ Ei ≤ 6.796 V
Assuming that eQ = ±Q/2 = ±1/2 LSB,
Ei = E* ± 1/2 LSB or 6.782 ≤ Ei ≤ 6.792 V
PROBLEM 7.21
KNOWN: A/D converter: M = 8; EFSR = 10 V; register = 1010 1011
FIND: Ei
SOLUTION
For two's complement code, 1010 1011 = -8510 [i.e. -(128 - 43)] which
is equivalent to
E* = (-85)Q = (-85)(10/28) = -3.320 V
Assuming that eQ = Q = 1 LSB,
Ei = E* + 1 LSB = -3.320 + 1 LSB or -3.320 ≥ Ei ≥ -3.359 V
Assuming that eQ= ± Q/2 = ± 1/2 LSB,
Ei = E*± 1/2 LSB or -3.339 ≤ Ei ≤ -3.300 V
PROBLEM 7.22
KNOWN: Dual slope A/D converter: M = 12; fclock = 10 kHz;
register = 201110
FIND tc
SOLUTION
The actual conversion time based on the register value of 2011 counts is
tc = (2)(2011)/10 kHz = 0.4022 s
PROBLEM 7.23
KNOWN: Single ramp A/D converter: M = 8; fclock = 1 MHz;
register = 17310
FIND tc
SOLUTION
The actual conversion time based on the register value of 173 counts is
tc = (173)/1 MHz = 173 µ s
PROBLEM 7.24
KNOWN: Successive approximation A/D converter: M = 8; EFSR = 10 V
Ei = 6.2 V
FIND: Register value
SOLUTION
The resolution is: Q = 10 V/28 = 0.039 V
So each register bit value counts as 39 mV.
For an input of 6.2V, the register count would be 6.2 V/0.039 V = 158.9.
If we account for register threshold errors, we take 158.9 ⇒ 15910 = 10011111 (straight
binary).
The comparator will see E* = 6.162 V after 7 steps and 6.201 V after 8 steps.
PROBLEM 7.25
KNOWN: Balance scale:
input range: 0 to 5 kg
output range: 0 to 3.50 mV
Recorder:
M = 12; EFSR = 10 V
FIND: Appropriate amplifier gain, G.
SOLUTION
The balance scale has a sensitivity: K = 3.5 mV/5 kg = 0.7 mV/kg
The recorder has a resolution: Q = 10 V/212 = 2.44 mV
It is clear that the recorder resolution is large compared to the balance output.
Amplification of the balance output prior to recording is in order. To improve
the recorder resolution to within, say, 1% of the scale full scale output we will
need, Q = (0.01)(3.5 mV) = 35 µ V. To achieve this,
G = 2.44 mV/35 µ V = 70
So, a minimum gain of about G = 70 is required. A higher gain will further improve
resolution.
COMMENT
Note that when an amplifier is used the resolution would be estimated as:
Q = EFSR/(G)(2M)
where G accounts for the gain.
PROBLEM 7.26
KNOWN: f ≈ 2 Hz
N δ t = 10s
Recorder: M = 12; EFSR = 10 V
FIND: fs, y(t), δ f , fN
SOLUTION
This problem is an open-ended design. One possible solution follows.
(i) To meet the Sampling Theorem, fs > 2fm. If we set fm = f = 2Hz, this gives fs > 4 Hz. But
because fm is not known exactly, we should set fs away from 4 Hz.
The minimum sample rate will also depend on N and the extent of spectral leakage (see
amplitude ambiguity). We should make some trial runs until we achieve acceptable results.
For a 10 s block of data, i.e. N δ t = N/fs = 10 s, so that with we have N = 10fs. As an
example, let fs = 20 Hz and N = 200.
(ii) y(t) = C1sin 2πft = 2 sin 4πt
[V]
(iii) From (i): δf = 1/Nδt = fs/N = 0.1 Hz; fN= fs/2 = 10 Hz. A 10 Hz low pass filter should be
used in front of the A/D converter.
1.6
dt=0.050 N = 200
1.4
AMPLITUDE
1.2
1.0
0.8
0.6
0.4
0.2
0.0
0
2
4
6
FREQUENCY [Hz]
8
10
PROBLEM 7.27
KNOWN: DAS: fs = 20 kHz; M = 12; Nδt = N/fs = 5 s
Computer: M = 16
FIND: Memory requirement
SOLUTION
Given that 5 s of data are acquired: N = (5s)(20,000/s) = 100,000
So 100,000 12-bit data points are acquired.
A 16-bit computer stores each integer number as a 16-bit word comprised
of 2 8-bit bytes. This requires that each 12-bit integer data point be stored as 2
bytes. So it takes 200,000 bytes to store the data. Now 1 kbyte = 1 kB = 1024
bytes.
So 200,000 bytes requires 195.3 kB of memory.
PROBLEM 7.28
KNOWN: DAS: M = 12; fs = 100 MHz
Computer: 8 MB of 32-bit memory
FIND: (Nδt)max
SOLUTION
Direct Memory Access (DMA) writes data directly from the acquisition process into
memory. If each 12-bit data point is stored as a 32-bit (4 8-bit bytes) word, then the
maximum storage is N = 8 MB = 8,192,000 bytes.
If fs = 100 MHz, then (Nδt)max= 0.08192 s.
If fs = 100 kHz, then (Nδt)max = 81.92 s.
COMMENT
When multiple channels are sampled, the number of data points acquired per unit time is
increased by the number of channels. The sample duration available is decreased by roughly
the number of channels (the amount is not exact because some time is required to flip
between channels). For example, a race car test might involve some 64 different sensors.
Even at 100 Hz, that is 6400 data points per second to acquire and then to analyze.
PROBLEM 7.29
KNOWN: Square wave: T1 = 1 s (Example 2.3)
FIND: N, fs for the sampling process. fc for the filter.
SOLUTION
This problem is an open-ended design. One possible solution follows.
We need to meet the criteria for Sampling Theorem and to minimize leakage (Amplitude
Ambiguity).
For the square wave of example 2.3,
y (t )= (4 / π ) sin(0.2π t ) + (4 / 3π ) sin(0.6π t ) + (4 / 5π ) sin(π t ) + (4 / 7π ) sin(1.4π t )
+(4 / 9π ) sin(1.8π t )
The first five non-zero terms show a frequency of fn = 2nπt/T1 for n = 1, 3, 5, 7, 9 with T1 =
10 s. So at n = 9, fm = 0.9 Hz. Hence: fs > 1.8 Hz.
The amplitudes are: Cn = 4/nπ for n = 1, 3, ..., 9.
To minimize leakage, we want mT1 = N/fs. This gives: N = 10mfs. Any combination of N and
fs will do provided fs> 1.8 Hz.
For example: fs = 3.2 Hz and N = 32m where m is a positive number. If we let m = 2 (i.e.,
sample for 2 periods of the signal), then we want N = 64.
For the ideal filter setting: set fc = fN = fs/2 = 1.6 Hz.
PROBLEM 7.30
KNOWN: Triangle wave: T1 = 2 s
FIND: Sampling process: fs, N. Filter: fc
SOLUTION
This problem is an open-ended design. One possible solution follows.
The first seven non-zero terms have the form:
y (t ) ≈ (4 D1 / π ) cos π t + (4 D1 / 3π ) cos 3π t + ... + (4 D1 /13π ) cos13π t
To meet the Sampling Theorem: For this series, fm = 13/2 = 6.5 Hz. So that fs > 2fm >13 Hz.
To minimize spectral leakage (amplitude ambiguity): We want mT1 = N/fs to minimize
leakage or N = 2fsm.
Any combination will work provided fs > 13 Hz. For example, fs = 16 Hz, N = 256.
An ideal filter is set at the Nyquist frequency: fc = fs/2 = 8 Hz.
PROBLEM 7.31
SOLUTION
This problem is an open-ended design. One possible solution follows.
One approach: let fs = 3.2 Hz, N = 64 (developed in Problem 7.29). The discrete Fourier
transform will return N/2 = 32 amplitudes corresponding to the N/2 frequencies spaced δf =
fs/N = 0.05 Hz apart and up to and including fN = fs/2 = 1.6 Hz. The time series is
y (t )= (4 / π ) sin(0.2π t ) + (4 / 3π ) sin(0.6π t ) + (4 / 5π ) sin(π t ) + (4 / 7π ) sin(1.4π t )
+(4 / 9π ) sin(1.8π t )
To generate a plot, we estimate the DFT using a spreadsheet or the accompanying software.
The 5- term series provides an
approximate square wave as
expected (see Chapter 2) of
the appropriate amplitude and
period.
5-term series
SIGNAL
1
0.5
0
-0.5
-1
-1.5
0
5
10
15
20
25
TIME [s}
1.4
dt=0.3125 N = 64
1.2
AMPLITUDE
The amplitude spectrum
returns 32 amplitude –
frequency pairings.
Amplitudes are zero, except at
the harmonic frequencies. It
shows the correct amplitude
Cn at each corresponding
harmonic frequency fn for n =
1,2,3,4,5. For example, C1 =
4/π at f1 = 0.1 Hz.
1.5
1.0
0.8
0.6
0.4
0.2
0.0
0
0.5
1
FREQUENCY [Hz]
1.5
2
PROBLEM 7.32
SOLUTION
This problem is an open-ended design. One possible solution follows.
Using the example from Problem 7.30: fs = 16 Hz, N = 256. The discrete Fourier transform
will return N/2 = 128 amplitudes corresponding to the N/2 frequencies spaced δf = fs/N =
0.0625 Hz apart up to and including fN = fs/2 = 8 Hz. D1 = 1 V.
y (t ) ≈ (4 D1 / π ) cos π t + (4 D1 / 3π ) cos 3π t + ... + (4 D1 /13π ) cos13π t
To generate a plot, we estimate the DFT using a spreadsheet or the accompanying software.
3
5-term series
2
SIGNAL
The 7- term series provides
an approximate triangle
wave (see Chapter 2)with
the appropriate amplitude
and period. The tips of the
triangle are somewhat
skewed with this short
series.
1
0
-1
The amplitude spectrum
returns 128 amplitude –
frequency pairs. Amplitudes
are zero, except at the
harmonic frequencies. It
shows the correct amplitude
Cn at each corresponding
harmonic frequency fn for n
= 1,2,3,4,5,6,7. For
example, C1 = 4/π at f1 = 0.5
Hz.
-2
-3
0
5
10
15
20
TIME [s}
1.4
dt=0.0625 N = 256
AMPLITUDE
1.2
1.0
0.8
0.6
0.4
0.2
0.0
0
2
4
6
FREQUENCY [Hz]
8
10
PROBLEM 7.33
KNOWN:
FIND:
RC filter
k=1
fc = 100 Hz
Attenuation at 10, 50, 75, and 200 Hz
SOLUTION
For a low pass RC Butterworth filter,
M(f) = 1/[1 + (f/fc)2k]1/2
where k = 1 for a single-stage filter. Note that with k = 1, the filter takes on the first-order
system form discussed in Chapter 3.
Recall that "attenuation" means reduction (Chapter 3). In the context of a filter, a device that
reduces the output amplitude (M(f) < 1) of targeted frequencies corresponds to a device with
a negative value of dynamic error at those frequencies. Dynamic error is given by
δ(f) = M(f) - 1
Note: Recall also that "gain" refers to a positive value of dynamic error.
f [Hz]
10
50
75
200
M(f)
0.995
0.894
0.800
0.447
δ(f)
-0.005
-0.105
-0.200
-0.553
attenuation [%]
0.5
10.5
20.0
55.3
PROBLEM 7.34
KNOWN:
FIND:
Low-pass LC Bessel filter with k = 3; fc = 100 Hz
Attenuation at 10, 50, 75, 200 Hz
SOLUTION
A three-stage, low-pass Bessel filter will have the transfer function (Chapter 6)
G ( s) =
ao
Dk ( s )
where Dk ( s ) =
(2k − 1) Dk −1 ( s ) + s 2 Dk − 2 ( s ) , Do ( s ) = 1 and D1 ( s )= s + 1 . For unit cut-off
frequency and with k = 3, D3(s) has the form
Dk(s) = D3(s) = (2k - 1)D2(s) + s2D1(s)
with D1(s) = s + 1 and D2(s) = s2+ 3s + 3. The three stage, unit cut-off
frequency transfer function is
G(s) = 15/[s3 + 6s2 + 15s + 15]
With fc = 100 Hz, let ωc = 2π f c = 628 rad/s. We substitute s / ωc for s and solve for s = jω
to obtain
=
G ( jω
) 15ωc3 /  − jω 3 − 6ω 2ωc + 15 jωωc2 + 15ωc3 
= M (ω )e jΦ (ω )
From which, the magnitude is
(
)
(
M (ω ) 15ωc3 /  (15ωωc2 − ω 3 ) 2 + (15ωc3 − 6ω 2ωc
=

f
10
50
75
200
ω
M(f)
%Attenuation (= |M(f) – 1|*100)
62.8
314
471
1256
1
0.98
0.94
0.63
0%
2%
6%
37%
)
2 1/ 2


PROBLEM 7.35
KNOWN:
FIND:
LC low-pass Butterworth filter
-3 dB ≤ M(0 ≤ f ≤ 5 kHz) and M(f ≥ 10 kHz) ≤ -30 dB
Values for R, C and k
SOLUTION
This problem is an open-ended design. One possible solution follows.
For a k-stage, low-pass Butterworth filter,
M(f ) =
1
1/ 2
1 + ( f f )2 k 
c


If we set f c = 5,000 Hz, then we meet the constraint -3 dB ≤ M(0 ≤ f ≤ 5 kHz)
immediately. To meet M(f ≥ 10 kHz) ≤ -30 dB, set the attenuation at10,000 Hz to be-30 dB.
Find the number of stages as
2k
Attenuation(
=
dB) 10 log 1 + ( f f c ) 


or
or solving to get k ≈ 5.
2k
−30db
= 10 log 1 + (10000 5000 ) 


A 5-stage filter has 5 reactive elements. To specify the capacitors and inductors, we refer to
Table 6.1 and its corresponding Figure. For a normalized value of Rs = 1Ω and a fc = 5000
Hz:
L = Li Rs / 2π f c
C = Ci / ( Rs 2π f c )
C1 = 19.7µF
C3 = 64µF
L2 = 103 mH
L4 = 103 mH
C5 = 19.7 µF
PROBLEM 7.36
KNOWN:
y(t) is a complex periodic waveform passed through a low pass
RC filter and then sampled discretely.
fs = 500 Hz
Filter cut-off: fc = 250 Hz
fm : The highest frequency harmonic of y(t) that can be sampled and keep
δ ( f ) ≤ 0.10
FIND:
ASSUMPTION
Sampling criteria selected for proper signal measurement.
SOLUTION
The input signal has the form
y (t=
) Ao + ∑ Cn sin(2π nf1t + Φ n )
n
We are looking for the value for fm = nf1 such that δ (=
f m ) M ( f m ) − 1 ≤ 0.10 .
M(f ) =
1
1/ 2
1 + ( f f )2 k 
c


where k =1 here. Setting fc = 250 Hz and M(fm) = 0.9, then
0.9 ≤
1
1/ 2
1 + ( f 250 Hz )2 
m


or f m ≤ 121 Hz.
With fc = 250 Hz, we know that M(250 Hz) = 0.707.
PROBLEM 7.37
KNOWN:
A 500 Hz component is to be removed from an analog signal and
passed through an A/D converter.
M = 8; EFSR = 10 V; fs = 200 Hz
FIND:
Low-pass anti-alias filter with M(500 Hz) ≤ eQ.
SOLUTION
A Butterworth low-pass filter will meet this need.
Quantization error: eQ = ± 0.5[EFSR/2M] = ± 0.5[10V/256] = ±0.01953 V
With fs = 200 Hz, an appropriate anti-alias filter will have fc ≤ 100 Hz.
We set fc = 100 Hz. To select the order of the filter,
M(500) = B/A = 1/[1 + (f/fc)2k]0.5 = 1/[1 + (500 Hz/100 Hz)2k]0.5
where the magnitude ratio is defined by the ratio of the output amplitude B to the input
amplitude A. The quantization error is fixed at 0.01953 V. The output amplitude at 500 Hz
depends on the input amplitude and M(500), or B = MA. Solving for k in terms of input
amplitude, A:
k = [2log A + 5.597]/1.3979
A [V]
k
0.02
0.10
1.00
10.00
2
3
4
6
So if A = 1 V, a two-stage filter will work; if A = 10 V, six stages are needed.
PROBLEM 7.38
KNOWN:
Sensor-transducer: Thermocouple
Thermocouple measures 50o ≤ T ≤ 70o C and the signal output is
2.585 ≤ E ≤ 3.649 mV.
Signal is digitized using an A/D converter:
M = 12; EFSR= 10 V (ie -5 to +5 V, bipolar); SNR = 40 dB
FIND: (a) eQ/E
(b) gain (G) required to reduce a to 5% or less
(c) estimate SNR in (b)
SOLUTION
(a) The quantification error of an M-bit device is estimate by its uncertainty due to its
resolution
eQ =
± EFSR / 2 M / 2 =
± (10V / 4096 ) / 2 =
± 1.22 mV
(
)
The relative quantification error would vary between
eQ/E = 1.22 mV/2.585 mV = ±0.472
or 47% at 50oC, to
eQ/E = 1.22 mV/3.649 mV = ±0.33
or 33% at 70oC. Both values are significantly large.
(b) One means to reduce this relative quantization error is through amplification of the
analog signal prior to quantization. To achieve 5% or less error requires an input signal of the
magnitude,
E = eQ/0.05 = 1.22 mV/0.05 = 24.40 mV
At 50oC (the smallest voltage quantized), this requires a linear amplifier gain of
G = Eo/Ei = 24.40 mV/2.585 mV = 9.44 ~ 10
Or roughly, use an amplifier having a linear gain of 10.
(c) Any signal is composed of a magnitude attributable to deterministic signal, Es and a
magnitude attributable to noise, En.
SNR = 20log Es/En
With SNR = 40 dB, Es= 100En.
COMMENT
During signal amplification, all information (within the frequency range of the amplifier) will
be scaled by the gain of the amplifier. So the SNR based on the incoming signal and signal
noise will not change. At the analog level, amplifying a signal will not affect the SNR. You
need a combination of filters and amplifiers to help with that.
But amplification of the analog signal occurs before quantization. Quantization error is added
during sampling. Quantization error will manifest itself as random noise on the sampled
signal. In effect, it reduces the SNR.
We see in part (a) that the sampled signal will be masked by this large quantization error
noise level. We see in part (b) that the signal level is raised to a level where it will not be
masked by the quantization noise. So a benefit in amplification is realized as a reduction in
the relative quantization error.
PROBLEM 7.39
KNOWN: A/D Converter: M = 8 or 12; EFSR = 10 V; 0 < fs ≤ 100 Hz
FIND: Specify M, fs, amplifier gain G, and filter (type, fc and k). Estimate eQ
and Nδt expected for your choice.
SOLUTION
These design problems have an open-ended solution path. To demonstrate, one possible
solution is presented for each case.
(a) E(t) = 2sin 20πt V
The input signal has an amplitude of A = 2 V and a single frequency of f1 = 10 Hz.
Accordingly, we will want
fs > 20 Hz
(1) fs > 2fm that is
(2) mf1 = Nδt m = 1, 2, ...
to correctly reconstruct both signal frequency and amplitude in the time discrete series. If we
set fs = 40 Hz so as to satisfy (1), we should sample the signal at data rates in multiples of 4,
i.e. from (2) N = 4m, m = 1, 2, ... . The sample period, Nδt, will then be consistent with (1)
and (2).
The 12-bit A/D converter is chosen for its better resolution and, hence, lower quantization
error:
Q = EFSR/2M = 10 V/4096 = 2.44 mV
eQ/E = ± 0.5Q/A = ±0.00122 V/(2 V) = ± 0.0006
Because A = 2 V, which fits within the ±5 V range of the converter, AND because the
relative quantization error is so small, no amplification is required.
An anti-alias filter is always required. A low-pass, LC Butterworth filter is selected for its
flat bandpass characteristics. Set fc = fs/2 = 20 Hz. The order of filter affects signal
attenuation at f1 = 10 Hz. The dynamic error is
δ(f) = M(f) - 1 = 1/[1 + (f/fc)2k]1/2 - 1
k
M(10 Hz)
δ(10 Hz) [%]
Attenuation [%]
2
3
0.971
0.992
- 2.99
- 0.8
2.99
0.8
Set k = 3 to provide an attenuation due to the filter of less than 1% at 10 Hz.
(b) E(t) = 1.5sin πt + 20sin 32πt - 3sin (60πt + π/4) V
The input signal contains amplitudes of A1, A2 and A3 with frequencies of f1 = 0.5, f2 = 16
and f3 = 30 Hz, respectively. The maximum frequency in the signal, fm, is 30 Hz. The signal
does not contain a single fundamental frequency. We need
(1) fs > 2fm or f > 60 Hz
(2) mT = Nδt
m = 1, 2, ...
Criterion (2) can be rewritten as m/f = N/fs. In order to meet both criteria for this signal
requires a minimum sample rate of fs = 240 Hz with N = 480 (i.e. Nf/fs = 1, 32 and 60, an
exact integer multiple of each of the three frequencies, respectively). Alas, the A/D converter
does not have such a capability! As one compromise, setting fs = 80 Hz with N = 160m will
meet the criteria for both the f1 and f2 components, but not for f3. We should expect leakage
about f3 in the discrete time series representation. An alternative approach is to use trial
and error on fs and N until leakage is reduced to acceptable levels.
The 12-bit A/D converter is chosen for its better resolution.
Q = EFSR/2M = 10 V/4096 = 2.44 mV
Amplitude A2 will saturate the A/D converter. We choose a linear amplifier with a gain of G
= 0.2 to keep the signal well within range. The relative quantization error becomes:
eQ/E = ± 0.5Q/GAi = ± 0.00122 V/GAi
which, for values of A1, A2, and A3, results in an eQ of 0.004, 0.0003, and 0.002 for the three
respective frequencies. Because this is below 1%, a value that we judge sufficient, the
amplifier gain seems appropriate.
An anti-alias filter is always required. A low-pass, LC Butterworth filter is selected for its
flat bandpass characteristics. Set fc = fs/2 = 40 Hz.
The order of filter affects signal attenuation at each fi . The dynamic error is
δ(f) = M(f) - 1 = 1/[1 + (f/fc)2k]1/2 - 1
k
M(0.5)
M(16)
3
5
7
1
1
1
1
1
1
M(30)
0.92
0.97
0.99
δ(30) [%]
-8
-3
- 0.9
Attenuation [%]
8
3
0.9
Set k = 7 to keep attenuation below 1% at all frequencies (see COMMENT).
COMMENT
A seven-stage (k = 7) filter is a high order for a passive filter circuit, but it will these
requirements. Note that the high k required results from the sample rate. If fs, and then fc,
were increased, the attenuation criterion could be met with a lower order filter. Try it!
(c) P(t) = -10sin 4πt + 5sin 8πt psi; K = 0.4 V/psi
The voltage signal sensed by the data acquisition system will be
E(t) = KP(t) = -4sin 4πt + 2sin 8πt V
The input signal has amplitudes A1 = 4 V and A2 = 2 V with f1 = 2 Hz and f2 = 2f1,
respectively. The maximum frequency is fm = 4 Hz. We want
(1) fs > 2fm or fs > 8 Hz
(2) mT1 = Nδt m = 1, 2, ...
Criterion (2) can be rewritten as m/f1 = N/fs. If fs = 10 Hz, then we should sample at data rates
of N = 2m.
The 12-bit A/D converter is chosen for its better resolution and small relative quantization
error:
Q = EFSR/2M = 10 V/4096 = 2.44 mV
eQ/E = ± 0.5Q/Ai = ±0.00122 V/Ai
For A1 = 4 V and A2 = 2 V, eQ/E = ± 0.0003 and 0.006, respectively. Because A1 and A2 fit
within the ±5 V range of the converter AND because the relative quantization error is so
small, no amplification is required.
An anti-alias filter is always required. A low-pass, LC Butterworth filter is selected for its
flat bandpass characteristics. Set fc = fs/2 = 5 Hz.
The order of filter affects the attenuation. The dynamic error is
δ(f) = M(f) - 1 = 1/[1 + (f/fc)2k]1/2 - 1
k
M(2)
M(4)
δ(4) [%]
1
5
7
0.98
1
1
0.78
0.95
0.98
- 22
- 5
- 2
We set k = 5 to keep attenuation below 5%.
COMMENT
One way to improve on this attenuation number without increasing the number of required
filter stages is to choose a higher value for fs. This enables selecting a higher fc. For example,
with fs = 50 Hz, N = 25m, we could set fc = 25 Hz requiring k = 1 to reduce attenuation well
below 1%.
PROBLEM 7.40
KNOWN:
y(t) = 4 sin 2πf1t + 2sin 2πf2t + 3 sin 2πf3t
where f1 = 4 Hz, f2 = 10 Hz, and f3 = 21 Hz
FIND: N and effective sample rate
SOLUTION
The signal frequency content is found by inspection. The maximum frequency is fm = 21 Hz.
To meet the Sampling Theorem criterion we will want
fs > 2fm = 42 Hz
To determine the sample size, the amplitude ambiguity criterion based on the fundamental
period T1 is mT1 = Nδt. This will eliminate spectral leakage.
4.5
dt=0.0156s N =
256
4.0
3.5
AMPLITUDE
This signal has no
fundamental period, so its
lowest frequency is used.
For f1 = 4 Hz,
T1 =1/4 s. So, N = m/4δt.
For m = 1 integer period: N
= 1/4δt = fs/4. Use some
reasonable combination of
these. For example, if we
want N to be an multiple
power of 2 (useful for many
DFT algorithms), then try N
= 256 and this sets fs = 64
Hz. This combination meets
both criteria. The resulting
DFT is exact.
3.0
2.5
2.0
1.5
1.0
0.5
0.0
0
5
10
15
20
25
30
35
FREQUENCY [Hz]
3.5
dt=0.009s N =
256
3.0
AMPLITUDE
Alternatively, a good
approximation of the signal
can be had by following the
rule to use an fs > 5fm.This
yields, fs > 5fm > 105 Hz.
Any N, fs > 105 Hz
combination works. Here we
try fs = 110 Hz and N = 256.
Note the spectral leakage
with amplitude degradation.
Try fs = 128 Hz and N =
256!
2.5
2.0
1.5
1.0
0.5
0.0
0
10
20
30
FREQUENCY [Hz]
40
50
60
PROBLEM 7.41
KNOWN:
Transducers: ±1 V output; ±25 cm H2O input
DAS: M = 10; EFSR = 10 V; 4 MB memory; 10 min. battery life.
T1 = 0.5 sec
FIND: N, fs, fc, G
SOLUTION
The transducer sensitivity is: K = 2V/50 cm H2O = 0.04 V/cm H2O.
The A/D resolution is: Q = 10V/210 = 0.00976 V. This can be expressed as Q = 0.00976
V/0.04 V/cm H2O = 0.244 cm H2O/bit.
Set G = 5. Although sensitivity already meets problem constraints, an analog amplifier
between the transducer and the A/D with a gain of G = 5 will take full advantage of the A/D
range and improve resolution:
Q = 10V/(5)(210) = 0.00195 V = 0.0488 cm H2O/bit.
So a full scale input on pressure (± 25 cm H2O) corresponds to full scale to the A/D (± 5V).
We are interested in mean values and signal dynamic frequency content at about f1 = 1/T1 = 2
Hz. So any fs > 4 Hz with appropriate anti-alias filter will meet the Sampling criterion. Using
the mT1 = N/fs criterion to minimize amplitude errors, we would try to set N = m(0.5)fs,
where m is an integer. At fs = 20 Hz, 12,000 samples will be taken per channel over a 10
minute period.
We should use an anti-alias filter set at fc = fs/2 .
COMMENT
You can always use higher sample rates. However, be aware that a higher sample rate means
a larger data set to analyze. In many cases, this is fine but sometimes it becomes a real
burden. In this example, an engineer at a race car track test might have 2 minutes to offer
advice on car modifications before the next test run. Being smart during acquisition will help
the engineer get the information faster (from analysis) so more time can be spent on
decisions.
Amp
Transducer
f
i
l
t
e
r
DAS
A/D
memory
card
Lay-out for Problem 7.41.
memory
card
PROBLEM 7.42
KNOWN:
Transducer in Problem 4.41 with accuracy of 0.25%
Measurement (precision) uncertainty of 0.10 cm H2O noted
FIND: ud at M = 10 and M = 12
SOLUTION
The transducer sensitivity is: K = 2V/50 cm H2O = 0.04 V/cm H2O
The A/D quantization uncertainty, uo, is estimated by
(uo)A/D = 10V/210 = 0.00976 V
or in terms of pressure
(uo)A/D = (0.00976 V)/0.04 V/cm H2O = 0.244 cm H2O
This just meets the requirements.
A typical good quality A/D converter might have an overall accuracy to within 1 LSB. Using
this value, the instrument error is
(uc)A/D = 0.244 cm H2O
The transducer has an instrument error of 0.25%. At full scale,
(uc)p = (0.0025)(25 cm H2O) = 0.0625 cm H2O
Then,
(ud)system = ± [ (0.244)2 + (0.244)2 + (0.0625)2]1/2 = ± 0.35 cm H2O
(95%)
If we include the additional knowledge that the data scatter adds a random uncertainty of
about 0.10 cm H2O, then this estimate becomes,
(ud)measurement = ±[(0.35)2 + (0.10)2]1/2 = ± 0.36 cm H2O
(95%)
Note how the A/D converter qualities dominate the overall uncertainty. The random
uncertainty from the measurement plays almost no role.
Improving the A/D converter to a 12-bit unit provides:
(u0)A/D = 0.061 cm H2O
Then, assuming an overall accuracy of 1 LSB,
(uc)A/D = 0.061 cm H2O
and
(ud)system = ± 0.11 cm H2O (95%)
Now, the transducer and A/D converter contribute more equally to overall uncertainty while
the uncertainty has been reduced notably.
Including the random uncertainty has an effect now, as
(ud)measurement = ± 0.15 cm H2O (95%)
PROBLEM 7.43
KNOWN:
Set-up of Figure 7.17
Ei = 3.333 V
Rs = R2 = R3 = R4 = 120 Ω
Rnull ≈39 kΩ
FIND:
Offset null voltage, Enull
SOLUTION
The adjustable trim pot has a nominal value at 39 kΩ. You adjust the value of the trim pot to
achieve the desired nulling voltage required to provide a zero output when the strain gauge is
not load (zero input). Consider the trim pot as the "fine tuning knob."
The offset null voltage available is found from
E

Ei R3 ( Rnull + Rs )
Enull =
± i −

 2 Rnull Rs + R3 ( Rnull + Rs ) 
 3.333V
(3.333V )(120Ω)(39000Ω + 120Ω) 
Enull =
±
−

(39000)(120) + 120Ω(39000Ω + 120Ω) 
 2
= ± 0.00256 V = ± 2.56 mV
Notice how this value will change: (1) if you use a different excitation voltage, (2) use a
different trim potentiometer resistance (often done by using a shunt resistor with the existing
trim pot).
COMMENT
The particular arrangement uses a single strain gauge sensor, known as a quarter-bridge
arrangement (as opposed to two strain gauge sensors, called a half-bridge, and four strain
gauge sensors, called a full-bridge).
The quarter-bridge sensor resistance is matched with a resistor of equal resistance, as indicted
by fixed resistor R3 = 120 Ω. R3 is sometimes called the "completion resistor." Because the
resistor resistance values around the bridge are nominal values (i.e., not exactly equal to
120 Ω), the bridge will likely not be balanced at zero input conditions. Hence, a nulling
voltage is used to subtract out any output voltage at zero input occurring due to slight
differences in resistor values.
PROBLEM 7.44
KNOWN:
FIND:
LC low-pass Butterworth filter
fc = 10 Hz
M(f = 5 Hz) ≥ 0.95
M(f = 20 Hz) ≤ 0.10
Rs = RL = 10Ω
Values for L, C and k
SOLUTION
This problem is an open-ended design and one possible solution follows.
For a low-pass, Butterworth filter such as shown in Figure 6.31, we begin by fixing the cutoff frequency (known to be 10 Hz here) and then estimating the order (number of reactive
stages) k needed to meet the constraints at 5Hz and 20 Hz. We end by specifying the C and L
values. For the filter:
M(f) = 1/[1 + (f/fc)2k]1/2
For f = 5 Hz, f/fc = 0.5; For f = 20 Hz, f/fc = 2
M(5Hz) = 0.95 = 1/[1 + (5/10)2k]1/2
This gives k = 1.6 ≈ 2.
M(20Hz) = 0.95 = 1/[1 + (20/10)2k]1/2
This gives k = 3.3 ≈ 4. Note: these equations can be manipulated to yield k directly as,
1 − M2
log(
)
M2
k=
2log(f/f c )
So select the higher order filter to meet both constraints: k = 4
A four-order filter has four reactive elements. To specify the capacitors and inductors, we
refer to Table 6.1 with its scaling equations:
C = Ci
1
2πf c R
L = Li
R
2πf c
Hence, for a normalized value of R = 10Ω and fc = 10 Hz, estimates for C and L yield:
C1 =
L2 =
1.218µF
0.294 H
C3 =
L4 =
2.94µF
0.122 H
PROBLEM 7.45
KNOWN:
Second order LC low-pass Butterworth filter (k =2)
fc = 100 Hz
FIND:
Attenuation at 10, 50, 75, 200, 400 Hz
ASSUMPTION:
Input and output impedance are 1Ω
SOLUTION
For a Butterworth low pass filter, the magnitude ratio is given by:
M(f) = 1/[1 + (f/fc)2k]1/2
In terms of decibels, this is dB = 20 log M.
Attenuation = 1 - M(f) or, in decibels, Attenuation(dB) = -20 log M
For f = 10 Hz, f/fc = 0.1
f
f/fc
M
10
50
75
200
400
0.1
0.5
0.75
2
4
0.9998
0.970
0.872
0.243
0.0623
Attenuation
0dB
0.26dB
0.6 dB
12.3 dB
24 dB
PROBLEM 7.46
Program Leakage allows the user to vary the sample rate and the sample number. It displays
the original and sampled waveforms and their corresponding amplitude spectra.
Use Leakage as a tutorial surrounding the discussion of Figure 7.4.
PROBLEM 7.47
Program Aliasing allows the user to vary the input frequency and the sample rate of a
continuous signal. It displays the original and sampled waveforms and their corresponding
amplitude spectra.
Use Aliasing as a tutorial to study the effect of sample rate on both the frequency content and
the time domain waveform. Look closely at the amplitude information in the sampled signal.
With this program, you can test the Sample Theorem and the Amplitude Ambiguity criteria
directly.
PROBLEM 7.48
Program Aliasing allows the user to vary the input frequency and the sample rate of a
continuous signal. It displays the original and sampled waveforms and their corresponding
amplitude spectra.
Use Aliasing as a tutorial to study the folding diagram (Figure 7.3).
fs = 10 Hz ⇒ fN = 5 Hz
5 Hz is displayed as the Nyquist frequency
So 5 Hz is the fN folding point on the folding diagram
0 Hz is the other folding point and 0 Hz is always one of the folding points.
Moving the slider on the signal frequency shows:
f = 10 Hz = 2fN ⇒ fa = 0 Hz
f = 15 Hz = 3fN ⇒ fa = 5 Hz
f = 20 Hz = 4fN ⇒ fa = 0 Hz
f = 8 Hz = 1.6 fN
f = 11 Hz = 2.1 fN
f = 13 Hz = 2.6 fN
f = 16 Hz = 3.2 fN
⇒
⇒
⇒
⇒
fa = 2 Hz
fa = 1 Hz
fa = 3 Hz
fa = 4 Hz
PROBLEM 7.49
Program Signal allows the user to vary the sample rate of a continuous signal (selected from
a pull down menu). It displays the original and sampled waveforms and their corresponding
amplitude spectra.
Use Signal as a tutorial to examine the effect of sample rate on both the time domain and
frequency domain discrete representations of the original signal.
PROBLEM 7.50
Program Leakage focuses on spectral leakage due to sampling. The user can select sample
rate and the number of sampled points so as to control the sample period. In this way, both
the length of the signal measured is controlled, as well as the spectral resolution δf.
As you vary
sample rate and
N, look at the
waveform
presented. This
is the
waveform
measured. In
particular, look
at the last
period
measured. If it
is a complete
period, there
will be no
leakage,
otherwise there
will be some.
Spectral
leakage is
eliminated
when the
dominant
frequencies are
centered within the interval define by each δf. This is the basis for the amplitude ambiguity
criterion. When a frequency is not centered within δf, some information spills out into
adjacent intervals.
PROBLEM 8.1
KNOWN/FIND: Define and discuss the significance of:
a) temperature scale
b) temperature standards
c) fixed points
d) interpolation
SOLUTION:
a) temperature scale - an established relationship for assigning numerical values
to measures of temperature. The absolute temperature scales are:
the Rankine scale, for U.S. customary units
the Kelvin scale, for SI units
b) temperature standards - a formally adopted and recognized means for practical
realization of temperature measurement. Standards provide a means for the
measurement of temperature which can be reproduced and agrees with the
thermodynamic definition of temperature.
c) fixed points - identifiable and experimentally reproducible conditions which are
associated with a certain temperature (numerical value). See Table 8.1.
d) interpolation - a method for determining temperatures other than those defined
by fixed points on a temperature scale. For the majority of applications, the
interpolating instrument is a platinum RTD.
PROBLEM 8.2
KNOWN: An apparatus to produce phase equilibrium points is required.
FIND: Describe the conditions necessary to establish phase equilibrium points. Identify
the effects of elevation, weather and material purity.
SOLUTION:
Other than the vapor-pressure-temperature points for helium and hydrogen, the fixed
points for ITS-90 are freezing points, melting points or triple points.
triple point - The procedure for calibrating a thermometer at a triple point is:
1. completely freeze the sample (an appropriate mass of material) in a closed
container
2. experimentally determine the energy required to melt the sample
3. re-freeze the sample
4. Add energy and record the thermometer output at 10, 20 40, 60, 70 and 80%
melted. These readings should agree.
This procedure demonstrates that a container capable of preventing contamination of a
sample of material, while allowing the removal and addition of energy, is required to
establish the triple point for a material. Similar requirements are needed for
melting/freezing points, with the notable exception that containers must generally be
flexible, to accommodate thermal expansion. Representative values of measured
temperatures agree to 0.1 mK.
A sample which is 99.9999% pure will produce measured temperatures over a phase
change within 0.1 mK.
Weather and elevation should be eliminated through appropriate design of the
experimental apparatus.
PROBLEM 8.3
KNOWN: A length of platinum wire having:
length, l = 2 m
diameter, D = 0.1 cm
resistivity, ρe = 9.83 × 10-6 Ω-cm
FIND: The resistance of the wire, R
SOLUTION:
Since
R=
and
ρe l
Ac
π
π
=
D2
=
( 0.1
) 0.0079 cm 2
4
4
The resistance is found as
=
Ac
2
( 9.83 ×10 ) ( 2 m ) (10
−6
R
=
0.0079 cm
COMMENT:
2
2
cm/m )
= 0.25 Ω
An RTD would normally have a reasonably large resistance, on the
order of 25 Ω. As such, a very small diameter wire or long length must be employed.
PROBLEM 8.4
KNOWN: A Wheatstone bridge and RTD as shown in Figure 8.35, with
α = 0.003925°C-1
Ro = 25 Ω at 0°C
R1 = 41.485 Ω for balanced conditions
FIND:
a) The temperature of the RTD
b) Compare the static sensitivity of this circuit
to the circuit in Example 8.2
SOLUTION:
At balanced conditions
R2 R1
=
R3 R4
and when R1 = 41.485 Ω
1=
From
41485
.
⇒ R4 = 41485
.
Ω
R4
[
R = Ro 1 + α ( T − T0 )
]
.
41485
= 25[1 + 0.003925( T − 0) ]
we find
41485
.
−1
25
and T = 168°C
T=
0.003925
b) The static sensitivities are the same, since R = R1 (R3/R2) and R2 = R3 in both cases.
PROBLEM 8.5
KNOWN: A thermistor has a resistance of 20,000 Ω at 100°C.
β = 3650°C
Ro = 20,000 Ω
R = 500 Ω
FIND: The temperature corresponding to a thermistor resistance of 500 Ω.
SOLUTION:
From (8.11)
1
Letting Ro = 20,000 Ω
R = Ro e
1

 T To 
β −
R 500
=
= 20, 000e
1 1 
3650 −

 T 373 
and
1 
1
ln 500 = ln 20, 000 + 3650  −

 T 373 
Solving for T
T = 598.7 K = 325.7°C
PROBLEM 8.6
KNOWN: The uncertainty in temperature uT =
±0.005°C .
FIND: Required uncertainty in measured resistance.
ASSUMPTIONS: Initially assume that we wish to find the required uncertainty in
resistance measurement as if it were the only contributor to the total uncertainty. In
addition, this problem is open-ended to some extent, in that some nominal value of Ro
must be assumed, or a range of values for Ro examined.
SOLUTION:
1
∂T
=
∂R αRo
Taking Ro = 100 Ω
With
[
R = Ro 1 + α ( T − To )
]
and α = 0.003925°C-1, we can express
uT =
∂T
u
∂R R
∂T
=
∂R
1
= 2.55o C/Ω
0.003925
(
)(100 )
and the uncertainty in resistance is
u R = ±0.00196 Ω
 R − 1
 R

o
 +T
Then T 
=
o
α
COMMENT: A parametric examination of the effect of the value of Ro on the
uncertainty would contribute to the fundamental understanding of the measurement (see
plot below). This is crucial at the design stage for a measurement system, and would
provide information concerning the sensitivity of the design to Ro.
Uncertainty, Ohms
Effect of Resistance on Uncertainty
0.004
0.003
0.002
0.001
0
0
50
100
150
Re s is tance , Ohm s
200
PROBLEM 8.7
KNOWN/FIND: Define and discuss the following terms related to thermocouple
circuits:
a) thermocouple junction
b) thermocouple laws
c) reference junction
d) Peltier effect
e) Seebeck coefficient
SOLUTION:
a) thermocouple junction - electrical connection between two dissimilar metals which form
a thermoelectric circuit.
b) thermocouple laws - observed behavior of thermoelectric circuits which allow the
measurement of temperature using thermocouple circuits.
c) reference junction - an emf is present in a thermoelectric circuit having two junctions
maintained at different temperatures. In order to measure temperature, one of the
junctions must have a known temperature, and is called the reference junction.
d) Peltier effect - this phenomenon results from the conversion of electrical to thermal
energy at a junction.
e) Seebeck coefficient - defines the relationship between temperature and emf for a
thermocouple circuit.
PROBLEM 8.8
KNOWN: For a J-type thermocouple, the measured emf at the potentiometer:
emf = 13.777 mV, for a reference junction temperature of 0°C
FIND:
Measuring junction temperature
ASSUMPTIONS: The J-type thermocouple is within NIST standards and Table 8.6
may be utilized.
SOLUTION:
From Table 8.6 with an emf = 13.777 mV and a reference junction temperature of 0°C
the temperature is found as 254oC.
PROBLEM 8.9
KNOWN: J-type thermocouple in Fig. 8.36 produces 15 mV for T1 = 750°C
FIND:
T2
SOLUTION:
The law of intermediate temperatures provides that
emfT − emfT = emfT −T
1
2
1
2
Referenced to 0°C, the emf corresponding to 750°C is 42.281 (Table 8.6)
Thus the reference junction temperature, T2, has an emf, referenced to 0°C of 42.28115 mV which yields 498°C.
PROBLEM 8.10
KNOWN:
T-type thermocouple in Fig. 8.36 produces 6 mV for T1 = 200°C
FIND:
T2
SOLUTION:
The law of intermediate temperatures provides that
emfT − emfT = emfT −T
1
2
1
2
Table 8.7 provides the reference function relating emf and temperature for a T-type
thermocouple referenced to 0°C. We must solve the following polynomial for T2:
n
n
6000 µV = ∑ ciT1 − ∑ ciT2
i =0
i
i
i =0
This equation is most conveniently solved in a spreadsheet or higher-level mathematical
software package. The value of T2 is 78.46°C.
PROBLEM 8.11
KNOWN:
a) Thermocouple circuit of Fig. 8.37a yields an emf
of 7.947 with Tref = 0°C
b) Tref = 25°C
c) Tref = 0°C with copper extension wires installed
FIND: The indicated temperature
ASSUMPTIONS: NIST standard thermocouple behavior
SOLUTION:
a) from Table 8.6, T = 148.9°C
b) Knowing emf1 + emf2 = emf3
7.947 mV = emf0-25 + emf25-148.9
with emf0-25 = 1.277 mV yields 6.67 mV for the output.
c) 148.9°C
PROBLEM 8.12
KNOWN: A J-type thermocouple referenced to 70°F.
output emf = 2.878 mV with Tref = 70°F.
FIND: The temperature of the measuring junction
ASSUMPTIONS: NIST Standard Behavior
SOLUTION:
To utilize Table 8.6, convert °F to °C
70°F = 21.1°C
and employing the Law of Intermediate Temperatures
emf0-21.1 = 1.076 mV
emf0-T = 1.076 + 2.878 = 3.954 mV
and
T = 75.7°C
PROBLEM 8.13
KNOWN: A J-type thermocouple referenced to 0°C;
output emf = 4.115 mV
FIND: The temperature of the measuring junction
ASSUMPTIONS: NIST Standard Behavior
SOLUTION: From Table 8.6 at an emf of 4.115 mV
T = 78.7°C
PROBLEM 8.14
KNOWN: An uncertainty level uT = 2°C at 200°C is required for a temperature
measurement using a T-type thermocouple. The readout device used for this temperature
measurement has:
accuracy: ± 0.5°C (e1)
resolution: ± 0.1°C (e2)
FIND: Determine if the uncertainty constraint is met.
ASSUMPTIONS:
NIST Behavior
SOLUTION:
The elemental errors associated with the indicator (output stage) may be combined as
e12 + e22 = 0.52 + 0.052 = ±0.503o C
This is the uncertainty that would result if the thermocouple exactly followed NIST
Standard Behavior. The uncertainty due to variations from the NIST Standard is found
from Table 8.5 as ±1.0°C or ± 0.75%, whichever is larger. This yields ±1.5°C, and
uT = 15
. 2 + 0.5032 = ±158
. oC
Yes, the uncertainty constraint is met.
PROBLEM 8.15
KNOWN: Thermocouple arrangement shown in Figure 8.21 with
N=3
J-type thermocouples
all junctions sense 3°C temperature difference
Maximum variation from NIST - 0.8%
Voltage measurement uncertainty ± 0.0005 V
FIND:
a) thermopile output for an average junction temperature of 80°C
b) the design stage uncertainty in measured temperature
SOLUTION:
The thermopile output will be 3 times that for 1 thermocouple sensing ∆T = 3°C at 80°C
(approximated from Table 8.6)
∂ emf
= 0.053 mV o C-1
∂T
The output is then
3 × (3°C) × (0.053 mV/°C) = 0.477 mV
b) First find the uncertainty in temperature which results from the uncertainty in the
voltage measurement
 ∂T 
uT = 
 uemf
 ∂emf 
But ∂T ∂emf is the slope of the emf vs T curve at 80°C for the thermopile. For a single
thermocouple, this slope is 1/0.053. For the thermopile, this slope is 1/[3(0.053)]
or 0.053 mV/°C. This yields
 1o C 
o
uT = 
 ( 0.05 mV ) = ± 3.1 C
 0.053 mV 
and there is a contribution due to the variation of the thermocouple from NIST standards,
which is related to the uncertainty in the slope of the curve shown below
At ∆T = 3oC, the
uncertainty in
temperature based
on the ±0.8%
yields ±0.024°C,
in the ∆T. Thus the resulting uncertainty is given by
uT = 0.024 2 + 31
. 2 = ±31
. oC
COMMENT:
The value of ∂emf ∂T is relatively insensitive to temperature; for example, at 400oC the
value is 0.055 mV/oC. The uncertainty in voltage of ±0.5 mV is unacceptable for most
temperature measurements, since the resulting uncertainty is higher than the measured ∆T.
However, if the number of junctions increased to 10, the resulting uncertainty would be
±0.94°C, which may be acceptable in many cases.
PROBLEM 8.16
KNOWN: Values of temperature and emf for a given reference temperature
FIND: Complete the table of values
ASSUMPTIONS: NIST Standard Behavior
SOLUTION:
Temperature [oC]
Measured
Reference
emf [mV]
100
0
5.269
-10
0
-0.501
100
50
2.684
96.6
50
2.5
PROBLEM 8.17
KNOWN: A thermopile having
4 junctions (N = 4)
Tref = 0°C
T = 125°C
uemf = ±0.0001 V = ±0.1 mV
FIND:
a) emf
b) N for an uncertainty of ±0.1 C
ASSUMPTIONS:
NIST Standard Behavior
J-type thermocouple
SOLUTION:
a) for a single thermocouple
emf1 = 6.634 mV
Thus for the thermopile the output would be
4(6.634 mV) = 26.536 mV
b) The static sensitivity of the thermocouple at 125°C is approximately 0.055 mV/°C and
 ∂T 
uT = 
 uemf
 ∂emf 
Thus
1
o
C mV ( 0.0001 × 10 3 mV)
N (0.055)
N = 18.2 or 19
0.1o C =
PROBLEM 8.18
KNOWN: A bimetallic thermometer serves as the sensing element in a thermostat for a
residential heating/cooling system.
FIND:
Considerations for
a) location for the installation of the thermostat
b) effect of the thermal capacitance of the thermostat
c) thermostats are often set 5°C higher in the air conditioning season
ASSUMPTIONS: Goal is to measure the air temperature inside the house
SOLUTION: Answers should address the following
1. Location should be on an inside wall to minimize conduction errors
2. Location should not be exposed to direct solar radiation, to prevent radiation
errors
3. Location should not be in the direct flow from the HVAC system
4. The thermal capacitance of the bimetallic thermometer typically yields a time
constant much shorter than required to regulate room temperature
5. Thermostats are typically set 5°C higher in summer primarily to save energy,
but also to accommodate seasonal lifestyle changes.
PROBLEM 8.19
KNOWN: A J-type thermocouple is to be used at temperatures between 0 and 100°C.
A single calibration point is available, at the steam point. Barometric pressure is
30.1 in. Hg, and the measured emf = 5.310 mV
FIND: Develop a calibration curve for this thermocouple
ASSUMPTIONS: (emfref - emfmeas) is linear from 0°C to 100°C
SOLUTION:
First, determine the steam point temperature for this barometric pressure
 30.1

 30.1

Tst = 212 + 50.422
− 1 − 20.95
− 1
 29.921

 29.921

which yields Tst = 212.30°F = 100.17°C
2
A calibration curve can be plotted with the dependent variable as (emfref – emfmeas) where
emfref = NIST standard emf value (mV)
emfmeas = measured output from thermocouple (mV)
emfref - emfmeas (µV)
Deviation Plot
0
-5
-10
-15
-20
-25
-30
-35
0
2
4
Measured emf (mV)
6
In this case, from Table 8.6
emfref = 5.278 mV
and we have a single data point at T = 100.17°C where (emfref – emfmeas) = - 0.032 mV
A second calibration point is known, since, at 0°C emfmeas = 0. Assuming linear behavior
for the error between these two points yields the calibration curve shown above. The
curve is used to correct a measured emf to an equivalent NIST standard thermocouple
output.
b) (Note: part b of this problem requires using some judgement in setting uncertainty
levels for various contributions to uncertainty)
One contribution to the uncertainty would result from the measured barometric pressure
(at the design stage)
 ∂T 
uT =   uP
 ∂P 
Assume up = ± 0.05 in. Hg and from
P

P

Tst = 212 + 50.422 − 1 − 20.95 − 1
 Po 
 Po 
2
∂Tst 50.422 2(20.95)  P 
=
−
 − 1
Po
Po
∂P
 Po 
at 30.1 in. Hg
∂Tst
= 1.677°F/in.Hg = 0.93°C/in. Hg
∂P
The contribution to uncertainty would be
uT = ( 0.93o C / in. Hg)(0.05 in. Hg) = ±0.047 o C
Some estimate of the uncertainty due to the assumed behavior of emfref – emfmeas must be
made. A reasonable estimate may be to take 1/4 of the maximum deviation, 32 µV, and
assign this value of uncertainty at the midpoint of the calibration range, such that in this
case at 2.5 mV the uncertainty would be
±8 µ V
yielding an uncertainty in temperature of ±0.00044°C
COMMENT:
Without additional measured data points, a reasonable estimate of the deviation from the
assumed linear behavior for emfref – emfmeas yields an uncertainty estimate. Engineering
judgement is required in applying this estimate for decisions in interpreting measured data
or in measurement system design or selection.
PROBLEM 8.20
KNOWN: A J-type thermocouple is calibrated against an RTD, yielding calibration data
over a range from 0°C to 100°C. The uncertainty in determining temperature using the
RTD is ±0.01°C over the range 0 to 200°C
FIND:
a) a polynomial to relate temperature and emf
b) the uncertainty in a measured temperature using the system as calibrated
c) the uncertainty in measured temperature using a specified indicator
ASSUMPTIONS: The calibration polynomial curve will be employed in data reduction
SOLUTION:
First, second, and third order polynomials for this data are
=
y 0.34058 + 18.833 x
y = 0.10989 + 19.157 x − 0.06075 x 2
−0.079403 + 19.926 x − 0.45169 x 2 + 0.048639 x3
y=
Choice of an appropriate polynomial can be made for a particular application, depending
upon the required uncertainty level. The standard error of the fit for the third order
polynomial is 0.34, and for the fourth order is 0.46.
b) Error contributions are
RTD - ±0.01°C
Potentiometer - 0.0012 + 0.0152 =
±0.015 mV
which is equivalent to a temperature uncertainty of
(
0.015 mV 1
0.055 mV/ o C
and the value of se taken to be 0.34, yielding
) = ± 0.27 C
o
uT =
0.012 + 0.27 2 + 0.342 =
±0.43 C
c) The readout uncertainty can be substituted for the potentiometer value and
uT =
0.012 + 0.322 + 0.342 =
± 0.47 C
PROBLEM 8.21
KNOWN: A thermocouple is placed in a moving gas stream with
U = 200 ft/sec
cp = 0.6 Btu/lbm R
h = 30 Btu/hr-ft2-R
Ts = 1200 R
Tp = 1400 R
r = 0.22
F=1
ε=1
FIND: a) T∞
b) er
SOLUTION:
a) the static temperature of the gas may be found from (8.37)
T∞ = Tp −
rU 2
2 gc c p
which yields
( 0.22)( 200) 2
T∞ = 1400 −
2( 32.174)( 0.6)( 778)
= 1400 R - 0.293 R = 1399.78 R
b) The radiation error may be found from (8.30)
hAs (T∞ =
− Tp ) FAsεσ (Tp4 − Ts4 )
( 30 Btu/hr ft R ) (T
2
T∞ = 1501.0 R
er = -101 R
∞
− Tp=
) 0.1714 ×10−8
Btu
14004 − 12004 ) R 4
2
4 (
hr ft R
PROBLEM 8.22
KNOWN: The static temperature of air outside an aircraft is to be measured.
U = 300 mph = 438.3 ft/sec
Altitude = 20,000 ft
r = 0.75
cp = 0.24 Btu/lbm-R
T∞ = 413 R
3
ρair = 0.0442 lbm/ft
FIND: Tp
SOLUTION:
From (8.36)
T=
T∞ +
p
= 413 +
rU 2
2 gcc p
0.75 ( 438.32 ) ft
2
sec 2
ft-lb m  
Btu  
ft-lb 

2  32.174
0.24
  778

2 
lb-sec  
lb m R  
Btu 

which yields
Tp = 425 R
PROBLEM 8.23
KNOWN: A sheathed thermocouple, as shown in Figure 8.38.
FIND: An estimate of the upper limit for conduction error for such a probe.
SOLUTION: From (8.27)
ec =
Tw − T∞
cosh mL
where
mL = hP kA L
For immersion depth as a parameter, an estimate of the conduction error requires a model
of the effective thermal conductivity of the thermocouple probe. A conservative estimate
for many constructions could be an average of the thermocouple, sheath and insulating
materials. Consider the following values:
kconstantan = 23 W/m-K
kstainless = 15 W/m-K
= 0.05 W/m-K
kinsul
Averaging these values yields 12.7 W/m-K. Considering the thermocouple probe to be
cylindrical in shape,
4h
mL =
L
keff D
For a probe having a diameter of 0.25 cm and an immersion length of 5 cm, the
conduction error is plotted as a function of h in the figure below.
Conduction Error
1
0.1
0.01
0.001
0.0001
ec /(Tw −T∞)
0.00001
0.000001
0
100
200
300
400
2
h [W/m K]
500
600
PROBLEM 8.24
KNOWN: An iron-constantan thermocouple is placed in a moving gas stream (as
shown in Figure 8.39)
Tref = 100°C
Tw = 260°C
h = 70 Btu/hr-ft2-°F
V = 200 ft/sec
emf = 14.143 mV
r = 0.7
cp = 0.24 Btu/lbm °F
ε = 0.25
FIND:
a) Tp
b) er and eu
ASSUMPTIONS: Radiation and velocity errors are additive
SOLUTION:
a) From Table 8.6, Tp = 355.8oC
b) The velocity error is given by
0.7 ( 200 )
rU 2
eU = Tp − T∞ =
=
= 2.33 R
2 g c c p 2 ( 32.174 )( 0.24 )( 778 )
and the radiation error by
2
εσ
0.25(0.1714 ×10−8 )
9604 − 1132.44 ) =
−4.87 R
er =
Tp − T∞ = (T − T ) =
(
70
h
4
w
4
p
The total error is then estimated as
e =eU + er =2.33 + (−4.87) =−2.54 R= − 1.4o C
PROBLEM 8.25
KNOWN: Ei = 1.564 V At 125°C, from example 8.5, BRT = 247 Ω
FIND: Show that BRT = 247 Ω, and determine the values of BRT at 150 and 100°C
SOLUTION:
Since
2
2
  ∂R
 ∂R
  ∂R

BRT =  T ( B) R1  +  T ( B) Ei  +  T ( B) E1 
 ∂R1
  ∂Ei

  ∂E1
and
( B )R
( B )E
=
±1.96 kΩ
2
=
±1.96 kΩ
i
1
The sensitivity indices are functions of temperature with
RT R1 
=

Ei
E1
− 1

and
Sensitivity Indices
∂RT  Ei

=
 E − 1
1

∂R1 
∂RT R1
=
∂Ei E1
∂RT − R1 Ei
=
∂E1
E12
100°C
0.042
125°C
0.022
150°C
0.012
86942
85238
84466
90591
87076
85505
This yields values of BRT of
T (°C)
100
125
150
BRT (Ω)
264
247
241
PROBLEM 8.26
KNOWN: A thermocouple circuit emf is measured by a potentiometer having limits of
error as
0.05% of reading + 15 µV at 25°C
and a resolution of 5 µV.
The connecting block temperature is 21.5 ± 0.2°C
and the potentiometer junctions are 25 ± 0.2°C.
FIND: T1
SOLUTION:
The error sources for the potentiometer may be combined,
2
=
uemf
 0.05

9000 ×15  + 2.52

 100

uemf =
±19.66 µ V =
±0.020 mV
Then since
 ∂T 
uT = 
 uemf
 ∂ emf 
and the sensitivity of the thermocouple is 0.055 mV/°C (from Table 8.6)
 1o C 
o
uT = 
 ( 0.020 mV ) = ±0.366 C
 0.055 mV 
The contribution from the uncertainty in the reference junction at the potentiometer is
±0.2°C, and the limits of error on the thermocouple are ± 2.2°C. Thus the total
uncertainty in temperature is
uT =
±2.24o C
( 0.366 ) + ( 0.2 ) + ( 2.2 ) =
2
2
The emf referenced to 0°C would be
emf0-T = 9mV + 1.096 mV = 10.096 mV
yielding
T = 187.7 + 2.24°C.
2
PROBLEM 8.27
KNOWN: A concentration of salt of 600 ppm in tap water will cause a 0.05°C change
in the freezing point.
FIND: Error in ice bath temperature having 1500 ppm of salt.
SOLUTION:
Consider two error sources for this ice bath,
1. Salt ± 0.125°C
2. Local temperature variations ± 0.05°C
The resulting design stage uncertainty may be found as
uT =
±0.135 C
( 0.125) + ( 0.05) =
2
2
T= 0 ± 0.135 C
PROBLEM 8.28
KNOWN: An RTD is to be calibrated; the RTD forms one leg of a Wheatstone bridge,
and has
=
α 0.00392 C-1 ± 1×10−5 (95%)
±0.001 Ω (95%)
uR =
At balanced conditions with T = 0°C, Rc =100.000 Ω and at 100°C, Rc = 139.200 Ω.
FIND: RRTD at 0°C and 100°C, and the uncertainty at the design stage at these
temperatures
SOLUTION:
At balanced conditions
RRTD Rc
=
Ra
Rb
Thus
RRTD = Rc
at 0C
=
RRTD 100.000 Ω
at 100C
=
RRTD 139.200 Ω
Expressing the relationship between temperature and resistance as
1  RRTD 
=
− 1 + To
T

α  Ro

the uncertainty at the design stage may be expressed, with γ =
RRTD
Ro
1
 ∂T 2  ∂T 2  2
=
uT 
uα  + 
uγ  
 ∂α   ∂γ  
The sensitivities are found as
∂T
=
∂α
∂T
=
∂γ
1− γ
α
2
 1− 3 
= 
= −130154 C2
2 
 0.00392 
1
= 255 C
α
at Rc = 300 Ω RRTD = Rc which implies RRTD = 300 Ω, Ro = 100 Ω γ = 3
We must estimate the uncertainty in γ . Since
RRTD =
Ra
Rc
Rb
with uR =
±0.001 Ω
uRRTD =
±0.003 Ω
( 3 × 0.001) =
and
1
2
2 2
 1
  − RRTD
 
=
uγ  uRRTD  + 
uRo  
2
 Ro
  Ro
 
1
2 2
2
 1
 
  −300

=
±0.0055
0.003  + 
0.001  =
2
  100
 100
 
yielding
1
( −130154 ×10−5 ) 2 + ( 255 × 0.0055 ) 2  2 =
uT =
±1.91 C


PROBLEM 8.29
KNOWN: A T-type thermopile is used to measure the temperature difference to
establish heat flux across an insulation. The pertinent variables and their values are:
Ac = 15 m2
L = 0.25 m
k = 0.4 W/m-K
∆T = 5°C
The uncertainty in the measured emf is ± 0.04 mV.
FIND: The number of junctions in the thermopile to yield an uncertainty level of 5% in
the heat flux across the insulation.
ASSUMPTIONS: NIST standard emf versus temperature relationship, and an
average temperature in the insulation of 40°C.
SOLUTION:
The heat flux is expressed as
∆T
L
For the purposes of the present analysis, express the uncertainty in Q as a percentage,
yielding
uQ u ∆T
=
Q
∆T
To determine the uncertainty in ∆T, the sensitivity to the uncertainty in emf must be
detemined. From the equation in Table 8.7, the value can be determined using
dE
= c1 + c2 T + c3 T 2 + c4 T 3 + c5 T 4 + c6 T 5 + c7 T 6 + c8 T 7
dT
Q = kAc
This expression yields a value of 0.042 mV/°C. Thus
u ∆T =
1
u
N ( 0.042) emf
where
uemf = ±0.04 mV
Then with
u ∆T
1
= 5% =
∆T
5N
which yields N = 4
PROBLEM 8.30
KNOWN: A T-type thermocouple referenced to 0°C is used to measure 100°C
FIND: The output emf.
ASSUMPTIONS: NIST standard emf versus temperature relationship.
SOLUTION:
From Table 8.7, the polynomial expression for emf as a function of temperature yields an
emf of 4.2785 mV at 100°C.
PROBLEM 8.31
KNOWN: A T-type thermocouple referenced to 0°C has an output of 1.2 mV.
FIND: The temperature of the measuring junction.
ASSUMPTIONS: NIST standard emf versus temperature relationship.
SOLUTION:
From Table 8.7, the polynomial expression for emf as a function of temperature yields an
temperature of 30.086°C for an emf of 1200 µV.
PROBLEM 8.32
KNOWN: A T-type thermocouple and voltmeter form a temperature measuring system
The temperature at the voltmeter is 25°C, and the output emf is 10 mV.
FIND: The temperature of the measuring junction.
SOLUTION:
The law of intermediate temperatures allows the following superposition to be used to
establish an equivalent emf referenced to 0°C.
emf 25−T + emf 0− 25 =
emf 0-T
From Table 8.7, the polynomial equation for E = f (T ) yields
emf 0-25 = 992 µ V
and
emf 0-T =
992 + 10, 000 =
10,992 µ V
which yields for T, from Table 8.7, a value of 231.542°C.
COMMENT: A calculator or mathematical analysis software is essential to solve for a
temperature from the polynomial expression in Table 8.7. NIST publications are also
available that contain tables of emf as a function of temperature for a variety of
thermocouple types.
PROBLEM 9.1
FIND:
Convert to units of gauge pressure in N/m2
SOLUTION
Absolute pressure reference scale:
1 atm abs = 14.69 psia = 101.325 kPa abs = 101,325 N/m2 abs
1 atm abs = 760 mm Hg abs = 406 in H2O abs
Conversion factors:
14.69 psi = 101.325 kPa = 101,325 N/m2
14.69 psi = 1 atm = 29.92 in Hg = 406 in H2O = 760 mm Hg
1 bar = 14.505 lb/in2 = 100,000 N/m2
p(gauge) = p(absolute) – p(reference)
Using p(reference) = 101,325 N/m2 :
(a) 10.8 psia x 101325 N/m2/14.69 psi = 74 442 N/m2 abs
p = 74,442 – 101,325 N/m2 = -26,883 N/m2 = -26.883 kPa
(b) 1.75 bars abs x 100,000 N/m2 = 175,000 N/m2 abs
p = 175,000 – 101,325 N/m2 = 73,675 N/m2 = 73.675 kPa
(c) 30.36 in H2O abs x 101,325 N/m2 /406 in H2O = 7,577 N/m2 abs
p = 7,577 – 101,325 N/m2 = -93,748 N/m2 = - 93.748 kPa
(d) 791 mm Hg abs x 101,325 N/m2 /760mm Hg = 105,458 N/m2 abs
p = 105,458 – 101,325 N/m2 = 4,133 N/m2 = 4,133 kPa
COMMENT
Note that gauge pressure can be negative or positive relative to the reference pressure.
Absolute pressure, which is measured relative to a perfect vacuum, is always a positive
value.
PROBLEM 9.2
FIND: Convert into absolute pressure
SOLUTION
p(absolute) = p(gauge) + p(reference)
where here: p(reference) = 1 atm abs.
Absolute pressure reference scale:
1 atm abs = 14.69 psia = 101.325 kPa abs = 101,325 N/m2 abs
1 atm abs = 760 mm Hg abs = 406 in H2O abs
Conversion factors:
14.69 psi = 101.325 kPa = 101,325 N/m2
14.69 psi = 1 atm = 29.92 in Hg = 406 in H2O = 760 mm Hg
(a) -0.55 psi + 14.69 psia = 14.14 psia = 0.963 atm abs
(b) 100 mm Hg + 760 mm Hg abs = 860 mm Hg abs = 1.13 atm abs
(c) 98.6 kPa + 101.325 kPa abs = 199.925 kPa abs = 1.97 atm abs
(d) 7.62 cm H2O + 760 mm H2O abs = 836.2 mm H2O abs = 1.10 atm abs
PROBLEM 9.3
KNOWN:
H = 250 cm H2O
patm = 101.3 kPa abs
FIND:
The tank pressure p1.
PROPERTIES: γH2O = 997 kg/m3
SOLUTION
Referring to Figure 9.5 (manometer)
p1 – p2 = p1 - patm
p1 = patm + γH2O H
= 101,325 N/m2 abs + (997 kg/m3 )(250cm)(1m/100cm)(1N/kg-m/s2)
= 103,818 N/m2 abs
or
= 0 N/m2 + (997 kg/m3 )(250cm)(1m/100cm)(1N/kg-m/s2) = 2493 N/m2
PROBLEM 9.4
KNOWN:
Deadweight tester provides a calibration pressure, pc
Tester:
W = 25.3 kgf = 2.58N (where gc = 9.8 kg-m/s2/kgf)
Ap = 5.065 cm2
Wp = 5.38 kgf
pamb = 770 mm Hg abs = 102.104 kPa abs
z = 20 m (sea level datum)
φ = 42o
FIND: pc
PROPERTIES:
γair = 9.8 N/m3
γstainsteel = γmass = 78.4 kN/m3
SOLUTION
Referring to the free-body diagram,
ΣFy = 0 = piAp + FB – Wp - W - pambAp
W
pambA
pamb(A – Ap)
Wp
piAp
FB
where FB is the buoyancy force. Neglecting FB, the indicated pressure is
pi = 122,313 N/m2 abs
Now, the actual pressure provided by the deadweight tester is estimated by
pc = pi(1 + e1 + e2)
where e1 provides a correction for altitude effects and e2 provides the correction for the
neglected buoyancy effects. From (9.6a),
e1 = - 0.0003
and from (9.9),
e2 = - γair/γmasses = -(9.8 N/m3 /78,000N/m3) = - 0.00012
Then, the actual deadweight pressure is estimated by
pc = 122,313 N/m2 (1 - 0.0003 - 0.00012) = 122,262 N/m2 abs = 122.262 kPa abs
COMMENT
By including the correction factors e1 and e2, we corrected for a known systematic error,
thus reducing the uncertainty error sources of this measurement. The error corrections do
not eliminate these systematic errors but do reduce them to the level of uncertainty in the
corrections themselves.
PROBLEM 9.5
KNOWN: Inclined tube manometer
∆L = 5.6 cm H2O
θ = 30o
FIND:
∆H
SOLUTION
Referring to Figure 9.7, for an inclined manometer,
∆H =
∆L sin θ
Further, this deflection away from a null balance condition is referenced to the pressure on
the open end of the tube. So the device measures a gauge (referenced to local atmosphere
pressure) or differential pressure (referenced to some other pressure).
∆H = 5.6 cm H2O x sin 30o
= 2.8 cm H2O
COMMENT
Referring to Figure 9.7, if the manometer fluid deflects towards the tank, the pressure
applied to the tank is below the reference pressure (p2<p1). If the deflection is away from
the tank, the pressure applied to the tank is above the reference pressure (p2>p1).
PROBLEM 9.6
FIND: Compare K for inclined and U-tube manometers.
SOLUTION
For a U-tube manometer, H is the measured output,
∆p= (γ m − γ ) H so
K = dH/d( ∆ p) = 1/( γ m − γ )
For an inclined manometer, H = Lsin θ where L is the measured output.
∆p= (γ m − γ ) L sin θ
so
=
K dL / d ∆=
p 1/[(γ m − γ )sin θ ]
PROBLEM 9.7
KNOWN: Inclined manometer measures air using mercury as its fluid.
θ = 30o
FIND: K
SOLUTION
For an inclined manometer, where L is the measured output
∆p= (γ m − γ )=
H (γ m − γ ) L sin θ
the sensitivity is
K dL / d ∆=
p 1/[(γ m − γ )sin θ ]
=
= 1/{[(13.6)(9800 N/m3) - 11 N/m3)]sin 30o} = 0.015 mm/N/m2
PROBLEM 9.8
KNOWN:
FIND: ud
Conditions of Example 9.2.
θ → 90o
p
SOLUTION
From Example 9.2:
(
)
1/ 2
2
2
2

±  uγ L sin θ + ( uL (γ m − γ )sin θ ) + ( uθ L(γ m − γ ) cos θ ) 
ud =
p
m


For a U-tube manometer at the stated conditions, L ≈ 10.25 mm and θ = 90o .
Then with γ m = 9770 N/m3; γ = 11.5 N/m3; uγ = 49 N/m3; uL = 0.0007 m
m
2 1/ 2
ud =
± 0.5 + 6.8 + 0 
p
2
2
=
± 6.82 N/m2
COMMENT
The uncertainty in measured pressure increases nearly 50% by going from an inclined
manometer with θ = 30o (Example 9.2) to a U-tube manometer ( θ = 90o) when operating
at these pressures.
PROBLEM 9.9
KNOWN: Steel diaphragm
t = 0.1 in
Em = 30 x 106psi
d = 2r = 0.75 in
ρ = 0.28 lbm/in3
ν p = 0.32
FIND: ymax, ωn , ∆pmax
SOLUTION
The maximum elastic deflection of a metallic diaphragm is about one third of the
diaphragm thickness,
ymax ~ t/3 = 0.033 in = 0.85 mm
The natural frequency can be computed directly
(30 × 106 lb / in 2 )(0.1in) 2 (32.2lbm − ft / s 2 − lb)(12in / ft )
Em t 2 g c
64.15
=
=
ωn 64.15
12(1 − ν 2p ) ρ r 4
12(1 − 0.322 )(0.28lbm / in3 )(0.375in) 4
= 2.8 x 106 rad/s or 450 kHz.
The maximum differential pressure which can be applied across a diaphragm is limited in
part by ymax. With ymax =
3( p1 − p2 )(1 − ν 2p )r 4
16 Em t 3
and ymax = t/3 gives
16(30 × 106 lb / in 2 )(0.1in) 4
∆pmax =
9(1 − 0.322 )(0.375in) 4
= 300,460 psi = 2.07 GPa
COMMENT
These relatively high numbers are due to the relatively thick and small diameter steel
diaphragm used. The numbers are not unusual for high pressure diaphragm transducers.
Some transducers on the market permit the user to interchange diaphragms of different
thicknesses to change the maximum pressure differential range allowed. This is a cost
saving feature for the user.
PROBLEM 9.10
KNOWN: Strain gauge, diaphragm tranducer with ∆p = 10, 100, 1000 kPa
Transducer:
Accuracy: within 0.1% of reading (i.e. u∆p/∆p = 0.001)
Voltmeter:
Resolution: 10 mV
Accuracy: within 0.1% reading
FIND:
ud in indicated pressure
ASSUMPTIONS: Transducer: Kt = 1 V/100 kPa ; Voltmeter: KE = 1 V/V
SOLUTION
Based on the assumptions for static sensitivity (you can assume any reasonable value), the
output voltage should be:
Eo = 1V @ 100 kPa, 0.1V @ 10 kPa, and 10V @ 1000 kPa
To estimate uncertainty:
Transducer:
u∆p = 0.001 ∆p
Voltmeter: (Note the inclusion of KE for scaling and units below)
uE = [uo2 + uc2]1/2 = [(0.005)2 + (0.001KEE)2]1/2
System: (Note the inclusion of Kt for scaling and units; that is, puts in terms of kPa)
ud = ±[u∆p 2 + (uE/Kt)2]1/2
With Kt = 1 V/100kPa and KE = 1 V/V,
∆p
[kPa]
10
100
1000
Eo
[V]
0.1
1.0
10.0
uE
[V]
ud
[kPa]
0.005
0.005
0.011
0.010
0.100
1.0
PROBLEM 9.11
KNOWN: U-tube manometer is used to measure gas pressure.
pgas ≤ 68,950 Pa
Several manometric fluids available: oil, water, mercury
FIND: Choose an appropriate manometric fluid.
PROPERTIES:
water: S = 1; γH2O = 9790 N/m3 (given)
mercury: S = 13.57 (given); γHg = SγH2O
oil: S = 0.82 (given); γoil = SγH2O
gas: γgas = 10.4 N/m3
ASSUMPTIONS: Specific weight of a gas is negligible relative to that of the
manometer fluids.
SOLUTION
To select an appropriate fluid, we must consider at least two things. (1) The manometric
fluid should not be soluble with the working fluid, and (2) the manometer deflection should
be of a reasonable magnitude. Neglecting the effects of the gases, the relation between
pressure and manometer fluid deflection is given by,
p - patm = ∆p = γ H
Water:
H = ∆p/γ = (68950 N/m2 )/9790 N/m2 = 7.1 m
(that's high!)
Oil:
H = ∆p/γ = (68950 N/m2 )/(0.82 x 9790 N/m2) = 8.6 m (that's high, too!)
Mercury:
H = ∆p/γ = (68950 N/m2 )/(13.57 x 9790 N/m2) = 0.52 m
(that's more like it!)
While sensitivity considerations will always favor the lighter fluid, the logistics of this
application clearly suggest that mercury will be a workable choice.
PROBLEM 9.12
KNOWN:
FIND:
Air pressure to be measured using either a mercury filled U-tube
manometer or inclined tube manometer.
200 ≤ p ≤ 400 N/m2
T = 20oC
U-tube Manometer:
Resolution: 1 mm; Zero error: 0.5 mm
Inclined-tube manometer
Resolution: 1 mm; Zero error: 0.5 mm;
Inclination angle: 30o ± 0.5o
ud in equivalent head pressure measured by either manometer
PROPERTIES: mercury: S = 13.57; water: ρ = 9780 N/m3
SOLUTION
U-tube manometer:
Because the measured deflection is the equivalent head pressure, the uncertainty in
equivalent head pressure at the design-stage will be due only to the ability to measure the
deflection at a given pressure.
ud = ±(uo2 + uc2)1/2
If we assume that uc is based only on the zero pressure error uncertainty given,
uo = 0.5 mm
uc = 0.5 mm
ud = ± (0.52 + 0.52)1/2 = ±0.71 mm
(95%)
This result is independent of pressure.
Inclined-tube manometer
The equivalent head pressure is related to the manometer deflection by,
H = L sin θ
where L is the measured deflection. We can estimate L:
L = H / sin θ = ∆p / γ sin θ
At 200 N/m2, L = 41 mm. At 400 N/m2, L = 82 mm.
For the inclined manometer, the uncertainty in the manometer deflection, L, at the design
stage is
1/ 2
1/ 2
ud =
± uo2 + uc2  = ± 0.52 + 0.52  = ±0.71 mm
L
where uo = 0.5 mm and uc = 0.5 mm as before. However, the uncertainty in equivalent head,
H, depends on the uncertainty in two variables, H = f(L,θ) , so
1/ 2
ud
H
 ∂H  2  ∂H  2 
uL  + 
uθ  
=
± 
 ∂L   ∂θ  
2
2
= ± ( sin θ uL ) + ( L cos θ uθ ) 


1/ 2
We set uθ = 0.5o = 0.0087 radians from the problem statement.
At 200 N/m2:
1/ 2
2
2
ud =
± ( 0.707 × 0.5mm ) + ( 41mm × 0.87 × 0.0087 ) 


H
= ±0.47 mm
(95%)
At 400 N/m2:
1/ 2
2
2
ud =
± ( 0.707 × 0.5mm ) + ( 82mm × 0.87 × 0.0087 ) 


H
= ±0.72 mm
(95%)
COMMENT
At the lower pressure, the uncertainty is reduced by a factor of about sin θ by using the
inclined manometer. But at higher pressures, the uncertainty in θ becomes increasingly
important. In this application, the uncertainty in θ cancels out the benefits of the better
sensitivity of the inclined instrument.
PROBLEM 9.13
KNOWN: A water filled inclined-tube manometer.
Inclination angle is variable.
∆p ≈ 10,000 N/m2
T = 20oC
Manometer:
Resolution: 1 mm; Zero error: 0.5 mm; Inclination error: ±1o
FIND:
ud in pressure as a function of θ
γ m = 9770 N/m3
PROPERTIES: water:
ASSUMPTIONS: uγ / uγ = 0.5%. Neglect effects of the ambient air.
m
SOLUTION
∆=
p L ( γ m − γ ) sin θ ≈ Lγ m sin θ
Then, ∆p =f ( L, γ m ,θ ) , so
u∆p
2
 ∂∆p
 ∂∆p


=
± 
uL  + 
uγ
 ∂L
  ∂γ m m

(
1/ 2
2
  ∂∆p  2 
uθ  
 + 
∂
θ

 


)
1/ 2
2
2
2

=
± ( γ m sin θ uL ) + L sin θ uγ
+ ( Lγ m cos θ uθ ) 
m


For the inclined manometer, the uncertainty in the manometer deflection, L, at the design
stage is
1/ 2
1/ 2
ud =
± uo2 + uc2  = ± 0.52 + 0.52  = ±0.71 mm
L
where uo = 0.5 mm and uc = 0.5 mm. From the problem statement
ud = (9770 N/m3)(0.005) = 49 N/m3
ud = 1o = 0.0175 rad
γm
θ
Then,
L
u∆p
[m]
[N/m2]
10o
30o
60o
80o
5.894
2.047
1.182
1.039
994
307
113
59
uθ dominates
uθ dominates
uθ dominates
uγ , uθ large
90o
1.024
50
uγ dominates
Inclination
m
m
COMMENT
At large pressures, uθ becomes very important (compare with Problem 9.12). In practice,
the inclination angle must be carefully set. But even so, the inclined manometer is selected
for deflections up to only about H = 0.25 m.
PROBLEM 9.14
KNOWN: Capacitance pressure transducer of Figure 9.14.
C1 = 0.01 ±0.005 µF
Ei = 5 ± 1% V
A = 8 ± 0.01 mm2
t = 1.5 ±0.1 mm
∆t = 0.2 mm
FIND:
C and Eo
SOLUTION
We know that C = cε A / t
and Eo =
C1
C
Ei
At to = 1.5 mm
C = (0.0885)(1)(8 mm2)/(1.5 mm)(10 mm/cm) = 0.0472 µF
and
Eo = (0.01/.0472)(5V) = 1.0593 V
At to + ∆t = 1.7 mm
C = 0.0416 µF
Eo = (0.01/.0416)(5V) = 1.2006 V
PROBLEM 9.15
KNOWN: Diaphragm pressure transducer:
uc = ± 0.5 psi
Voltmeter
uc = ±10 µV
uo = 1 µV
System (0 to 100 psi)
p = 0.564 + 24E ± 1 psi (95%) with N = 5
Installation errors: B = ±0.5 psi
FIND: up
SOLUTION
The system sensitivity is: K = dp/dE = 24 psi/V. The system uncertainty includes,
Instrument errors, uc:
transducer uncertainty ut = ±0.5 psi
voltmeter uncertainty uE = (1x10-6 V)(24 psi/V) = ± 2.4x10-4 psi
These instrument errors are normally treated as systematic errors or Type B errors.
Data reduction errors, u1:
curve fit uncertainty uyx = ± 1 psi with ν = 4
The curve fit error is usually treated as a random error or Type A error.
Installation effects, u2:
installation errors uin = ± 0.5 psi
Installation errors are usually treated as systematic errors or Type B errors.
So that: up = ± [.52 + (2.4x10-4)2 + 12 + .52]1/2 = ± 1.22 psi (95%)
Alternatively, we can segregate the errors into types and developed the expanded
uncertainty:
ut = Bt; uE= BE; uin = Bin; uyx= tPyx = tSyx/N1/2 with ν = 4 then,
B=
(∑ B )
2
k
1/ 2
and P = ( ∑ Pk2 )
1/ 2
so that
up = ± [B2 + (tP)2]1/2 = ± 1.22 psi (95%)
COMMENT
There is no magic in uncertainty analysis. Different approaches will yield similar (but not
necessarily identical) results provided that the same errors are included. The important
thing is to do an analysis!
PROBLEM 9.16
KNOWN: pressure transducer: t90 = 10 ms; ωd = 200 Hz; ζ = 0.8
FIND: Test plan to verify given specifications. Estimate frequency response.
SOLUTION
This problem is open-ended and could form the basis of a lab exercise. One solution is
discussed. Second-order systems are discussed in detail in Chapter 3.
(i) A step test should be developed to test for the rise time. An appropriate magnitude for
the pressure rise is the step from atmospheric pressure to the expected pressure at top dead
center for this engine (note: the compression ratio for an IC engine is about 8:1 or 9:1).
Transducer output can be measured on a storage oscilloscope or suitable data acquisition
system.
For a damping ratio of 0.8, the transducer will still exhibit a modest ringing during a step
test. But it will require excellent resolution in the measuring device to observe and to
accurately measure the amplitude changes. With adequate resolution, the maximum
amplitudes in the oscillation can be plotted versus time to estimate the product ωnζ (note:
this will be the slope of the line if plotted on semi-log axes). The period of oscillation will
be related to=
ωd ωn 1 − ζ 2 . The damping ratio and natural frequency are found by
solving these two pieces of information simultaneously.
(ii) Car with 4-cylinder engine turning at 5000 RPM = 83 rps. This gives expected pressure
changes at about 83/4 ~ 21 Hz. For the transducer,
=
ωn ωd / 1 − ζ 2 or
=
f n f d / 1 − ζ 2 = 200 Hz/(1 - .82)1/2 = 333 Hz.
Checking out the magnitude ratio plot for a second order device: f / f n ≈ 0.063 and
M(21Hz) ≈ 1. So yes, it could measure the pressure variations in the engine.
PROBLEM 9.17
KNOWN: Steel diaphragm transducer: t = 0.001 m; r = 0.003 m
FIND: ωn , pmax
SOLUTION
The natural frequency is given
ωn = 64.15
Em t 2 gc
12(1 − ν 2p ) ρ r 4
= 173,000 r/s
= 64.15
200 × 109 N / m 2 × (0.001m) 2 × 1kg − m / s 2 − N
12(1 − 0.352 )(7832kg / m3 ) × (0.003m) 4
or fn = 127 kHz
The maximum elastic displacement of the diaphragm is limited to about t/3.
=
∆p 16 Em t 3 (t / 3) / 3(1 −ν p2 )r 4
For steel with Em = 200x109 N/m2, ρ = 7832 kg/m3 and ν p = 0.35:
∆p = 16 × 200 × 109 N / m 2 × (0.001m)3 (0.001m / 3) / 3(1 − 0.352 )(0.003m) 4
= 5000 MPa
This is a very high pressure limit, but reflects the stiffness of steel and the relatively small
radius, thick diaphragm used. A larger radius for a given thickness lowers the natural
frequency and the maximum pressure: For example, doubling the radius here lowers ∆p by
16 times (r4) to about 312 MPa.
PROBLEM 9.18
KNOWN: Pressure transmission system filled with air at 20oC
L = 0.25 m
d = 3.25 mm
∀= 1600 mm2
Transducer:
fn = 100kHz
FIND: ωmax such that 0.9 ≤ M(ω) ≤ 1.1
PROPERTIES: Air: µ= 1.8 x 10-5 N-s/m2 ; R = 0.287 kJ/kg-K
ASSUMPTIONS: Air behaves as a perfect gas
SOLUTION
2
=
∀t π d=
L / 4 π (3.25mm)2 (250mm) / 4 = 2400 mm3
With ∀t > ∀,
ωn =
ζ =
a
L(0.5 + 4∀ / ∀t )
16 µ L 0.5 + 4∀ / ∀t )
ρ ad 2
If we assume that the process occurs at a pressure near atmospheric pressure, then we can
compute the density of the air by
ρ= p/RT = (101325 N/m2 abs)/(0.287 kJ/kg-K)(293 K) = 1.16 kg/m3
The speed of sound (acoustic wave speed) is
a = [kRT]1/2= [(1.4)(0.287 kJ/kg-K)(293 K)]1/2 = 345 m/s
Then,
ωn = 775 rad/s
ζ = 0.026
The frequency response is
M (ω ) =
1
[1 − (ω / ωn ) ] + [2ζ (ω / ωn )]2
ωn [rad/s]
10
100
200
225
250
300
2 2
M( ω )
1.00
1.02
1.07
1.09
1.12
1.17
The frequency response remains within the ±10% constraint over the frequency band,
0 ≤ ω ≤ 230 rad/s.
COMMENT
Note how the natural frequency of the tubing is much less than that of the transducer. As a
consequence, the response characteristics of the connecting tubing govern the system
response. The limits of the frequency response of the transducer do not come into play.
The assumption concerning the pressure affects the density of the air only. Its effect on the
solution is minimal for low gauge pressures.
PROBLEM 9.19
KNOWN:
Pitot static tube used to measure velocity
FIND: Sensitivity of velocity to pressure
SOLUTION:
A pitot-static tube, such as shown in Figure 9.23, can be used to determine velocity based
on the measured difference in total and static pressure through
U = 2∆p / ρ = C ∆p
where C is a coefficient.
The sensitivity is the change in output for a change in input or
K=
∂U 1
= C∆p −1 / 2
∂∆p 2
So the static sensitivity here is inversely related to the pressure. That is, the sensitivity of
the pitot static tube to changes in velocity decreases as the pressure difference increases.
COMMENT
The pitot-static principle is widely used and considered a fairly accurate means of
estimating velocity. This is a good example of a useful device whose sensitivity is not a
constant but varies with the magnitude of the input signal.
PROBLEM 9.20
KNOWN:
FIND:
Pitot-static probe measures flow in a duct.
H = pv / γ H 2O = 20.3 cm H2O
U
ASSUMPTIONS: Duct flow is air at room temperature.
PROPERTIES: ρ air = 1.16 kg/m3
ρ H 20 = 998 kg/m3
SOLUTION
1
ρ airU 2
2
= 2γ H 2O H / ρ air = 2 ρ H 2O gH / ρ air
pv = pt − p =
U = 2 pv / ρ air
= 2(998kg / m3 )(9.8m / s 2 )(0.203m) /(1.16kg / m3 )
= 58.5 m/s
PROBLEM 9.21
KNOWN: Pitot-static tube in air.
FIND: U', uU
SOLUTION
The mean velocity is determined by converting the mean voltage values to pressure and,
finally, to velocity. From the given data, the pooled estimates are:
E = (2.438 + 2.354 + 2.473)/3 = 2.422 V
<SE> = [(.012 + 0.0092 + .0122)/3]1/2 = 0.01 V
ν E = 60
From the calibration data, p = f(E), so we determine:
p = 0.205 + 0.950E = 2.506 N/m2
Now variations in voltage are related to variations in pressure by the static sensitivity at the
operating voltage (the mean voltage here), as noted in discussions related to Figures 1.6 and
5.2. Here Kp = dp/dE = 0.95 N/m2/V.
<Sp> = Kp<SE> = 0.0095 N/m2
with ν E = 60
The velocity is estimated by U = 2 pv / ρ air where pv is the dynamic pressure measured by
the pitot-static tube sensor. The pooled mean estimate is
=
U
2(2.506 N / m 2 )(1kg − m / s 2 − N ) /(1.2kg / m3 ) = 2.04 m/s
Variations in pressure are related to variations in velocity. Here KU = dU/dp =
= ( 2 pv / ρ air )
−1/ 2
= 0.41 m/s/N/m2 at pv = 2.506 N/m2:
<SU> = KU<Sp> = KUKp<SE> = 0.004 m/s
This value represents the effect of the variations in the test data during repetition. The effect
of variation in the measured mean value during replication is estimated by:
SU = KUKpSE = KU K p  ∑ ( E ′j − E ) 2 / 2 


1/ 2
= 0.022 m/s with ν = 2
For the voltmeter: an systematic uncertainty due to instrument error is assigned
B1 = ±10µV = ±10µV (.95 N/m2/V)(0.41 m/s/N/m2) = 4x10-6 m/s
P1 = 0
For the pitot-static tube, a systematic uncertainty due to instrument error of 1% is assigned:
B2 = (0.01)KUpv = (0.01)(0.41 m/s/N/m2)(2.506 N/m2) = 0.010 m/s
P2 = 0
For the transducer: no information is given, so assume a negligible instrument error.
B3 =0 and P3 = 0
From the calibration data tSyx/N1/2 = 0.002 N/m2 with ν = 30, so Syx/N1/2 ~ .001 N/m2 (i.e.
t~ 2)
P4 = (0.001 N/m2)(.41 m/s/N/m2) = 0.00041 m/s
B4 = 0
For the measurement:
ν 4 = 30
P5 = 0.004 m/s with ν 5 = 60 (repetition effect)
B5 = 0
From the replication study,
P6 = SU = 0.022 m/s
B6 = 0
(replication effect)
Collecting terms:
B = (B12 + B22 + B32)1/2 = 0.010 m/s
P = (P42 + P52 + P62)1/2 = 0.024 m/s with ν = 3 (from Welch-Sattertwaite method)
so,
uU = [B2 + (t3,95P)2]1/2 = 0.07 m/s
U' = 2.04 ± 0.07 m/s
(95%)
PROBLEM 9.22
KNOWN:
Pitot-static tube in freestream
U = 325 kph
(a) measured on open road
(b) measured in open circuit tunnel (where pt = 0)
FIND: Static, dynamic, stagnation pressures
SOLUTION
For a pitot-static tube in air,
U = 2∆p / ρ air
Rearranging, the dynamic pressure is given by
∆p =
1
ρ U2
2 air
so,
∆p = (0.5)(1.225 kg/m3)(325 km/hour x 1 hour/3600 s x 1000m/km)2
= 4991 N/m2 = 37.4 mm Hg
This is the measured pressure. Now, ∆p = pt – pstatic, where pt is the stagnation pressure and
the static pressure pstatic is the freestream pressure (p∞).
On the open road:
In this case, the flow is at rest and the car (and pitot-static tube) moves through the air,
causing the pressure to rise at the impact port (stagnation pressure).
p∞ = 0 N/m2 (local atmospheric pressure)
pt = 4991 N/m2
In the open circuit wind tunnel:
In an open circuit wind tunnel, air is drawn from rest from outside the tunnel and
accelerated to a desired speed within the test section, which houses the car.
The car is at rest and the flow is moved past it. Flow always moves from high to low
pressure, from the atmospheric pressure outside the tunnel to the low pressure inside it. The
pressure in the test section is the freestream pressure and clearly this must be below
atmospheric pressure for air to move through the test section. The stagnation pressure in the
tunnel is zero (just the same as it is for atmospheric pressure). Another way to think of the
stagnation pressure is that since the local pressure is negative, the pressure at the impact
port must rise an amount to reach atmospheric pressure.
p∞ = - 4991 N/m2
pt = 0 N/m2
PROBLEM 9.23
KNOWN: Wall pressure taps connected by tubing to transducers.
FIND: Select tube size for better time response.
SOLUTION
The system is depicted in Figure 9.21. From equation (9.27):
(L ) 2
L
L2
∝ d
τ∝ 4 =
2
2
∀
d
( Ld )d
or the system time constant is proportional to L/d per unit volume. So this indicates to keep
τ small that it is best to keep lengths as short as possible and diameters of moderate size,
but certainly not very small.
In fact, very small diameter tubes can be used as effective filters to suppress higher
frequency aspects from a pressure signal. This possibility is seen in equations (9.21 –9.26)
in which natural frequency decreases as L increases and damping rises as diameter
decreases.
PROBLEM 9.24
KNOWN:
System of Example 9.7
Average pressures and deviations.
FIND: Uncertainty in pressure measurement
SOLUTION
This problem offers a realistic scenario and requires some straightforward thought.
The resolution of the transducer and A/D system is limited to
Qtransducer = 50.8 cm H2O/5V = 10.16 cm H2O/V
QA/D = 5V/212 = 0.00122 V/digit
Or, the total measuring system resolution is
Q = 0.00122*10.16 = 0.0124 cm H2O
This yields a uo = 0.0124 cm H2O. The accuracy of the A/D is 2 bits, equivalent to
uc = (2 x 0.00122V)(10.16 cm H2O/V) = 0.025 cm H2O.
The uncertainty in the A/D signal is:
uA/D = [uo2 + uc2)1/2 = 0.028 cm H2O
In the wind tunnel the measurements will be sampled over a reasonable period of time to
acquire a large data set. Uncertainty enters the estimation of the average pressure due to
variations in the measured data. Some locations on the car will show larger variations than
do others. Based on the sample data given, the variation in pressure is on the order of
0.0025 to 0.05 cm H2O. Assuming large data sets (such that t95 ≈ 2.0), the uncertainty in
pressure due to variations alone is 0.005 to 0.10 cm H2O. The uncertainty in estimating the
mean pressure will be about one order of magnitude better than this, 0.0005 to 0.01 cm
H2O.
So our ability to resolve pressure is limited at one extreme by the A/D system at 0.028 cm
H2O. Our precision in a pressure measurement is reasonably estimated at 0.10 cm H2O and
provides the other extreme. The precision in the mean pressure estimation is set at 0.01 cm
H2O. Thus,
up = [.12 + 0.0282]1/2 ≈ 0.10 cm H2O (95%)
and
u p = [.012 + 0.028 2 ] ≈ 0.03 cm H 2 O (95%)
COMMENT
To put this in perspective of Example 9.7, this uncertainty level provides an uncertainty in
rear down force estimation of about 10 lbs (2.2 N) out of 600 lbs (135 N)) when done
correctly. Incredibly, this is about the level when a top professional driver can begin to
sense changes in car handling.
PROBLEM 9.25
KNOWN: Pressure measured at M = 4 stations, N = 20 each.
FIND: Effects of data pooling
SOLUTION
Pooling of the data obtained at 4 different radial planes would provide an overall average
pressure which accounts for non-axisymmetric effects, that is spatial variations between
cross-planes. In fact, by comparing the mean values found along each plane, an estimate of
the non-symmetry can be found.
p = (153 + 142 + 161 + 157)/4 = 153.25 MN/m2
A measure of the average variation along any cross-section is estimated by
<S> = [(72 + 92 + 92 + 72)/41/2 = 8.1 MN/m2
A measure of the spatial variation in the mean is inferred by (with M = 4)
(
M
Sp =
∑ p j − p
 j=1
)
2
1/ 2

/(M − 1) 

= 8.2 MN/m2
Note: The closeness of <S> and S here is just coincidence of the problem numbers.
Replications provide a means to estimate how well the test conditions and their effect on
the measured results can be repeated.
PROBLEM 9.26
KNOWN:
Air flow measured using a pitot-static tube and a mercury filled
manometer.
5 ≤ U ≤ 50 m/s
FIND: Manometer resolution required to achieve a zero order uncertainty of
5 and 1% in measured velocity.
SOLUTION
The velocity of a flow of fluid of density ρ bringing about a manometer deflection H is
given by
U=
2 γ Hg H / ρair =
2ρHg gH / ρair
or
1
H=
( ρair U 2 ) /(ρHg g)
2
To assess the effect of uH on uU :
1/ 2
2

 
 ∂U

 1 −1/ 2 ρHg g  
uU =
u   =
u
± 
± H
2
ρair H  
 ∂H H  



 

dividing through by the expression for velocity U
1/ 2
2
u U / U = ± u H / 2H
So the relative uncertainty in velocity is one-half the relative uncertainty in manometer
deflection. The uncertainty at any level can be found by inserting the corresponding and
consistent value of uncertainty into this expression. At the zero order, our concern is only
on the resolution of the measuring instrument. For example, for uUU = 0.01, uH/H =
± 0.02 , or we need a manometer resolution of 4% of the expected manometer deflection.
U [m/s]
5
50
H [mm Hg]
0.115
11.49
Resolution [mm]
1%
5%
0.00046
0.023
0.46
2.3
The lower velocity readings require micron level resolution if using mercury.
PROBLEM 9.27
KNOWN: Static pressure around cylinder, p(θ) .
pv = 20.3 cm H2O
p∞ = patm = 101.3 kPa abs
T∞ =Tatm = 16oC
Manometer measures pressure using deflection of water
FIND: U( θ )
ASSUMPTIONS: Total pressure constant throughout tunnel (i.e., neglect losses).
PROPERTIES: Water: ρ = 1000 kg/m3
Air: ρ = p/RT = 1.2 kg/m3
SOLUTION
For no losses, the Bernoulli equation can be written along a streamline from the freestream
to the stagnation point as
p t = p∞ + 0.5ρU ∞2
Similarly, we can relate to the pressure around the body’s surface
p∞ + 0.5ρU ∞2 =p(θ) + 0.5ρU 2 (θ)
so that,
p t = p(θ) + 0.5ρU 2 (θ)
Then, rearranging we get the velocity as a function of measured pressure
2(p t − p(θ))
U(θ) =
ρair
The manometer will deflect relative to the pressure difference, p(θ) − p∞ . But at the 0o tap,
p(θ) =p t , so the manometer will deflect to the pressure difference, p t − p∞ . In terms of
deflection,
H(=
θ 0o=
) (p t − p∞ ) / γ H2O
But we see that H(=
θ 0o=
) 0 , meaning that pt = p∞. This allows us to rewrite the
expression for velocity as (where H [cm] and U [m/s]),
=
U(θ)
2(p∞ − p(θ))
=
ρair
2ρH20 gH
= 12.65H1/ 2
ρair
θ
0
45
90
135
180
H [cm H2O]
0
41.4
81.3
23.1
23.9
U [m/s]
0
81.4
114.0
60.8
61.8
PROBLEM 9.28
KNOWN: Pitot-static probe. Tamb = T∞ =20 oC
FIND: Lowest airspeed and manometer deflection for which the viscous
correction is negligible.
PROPERTIES: Air: ρ = 1.225 kg/m3
ν = 2 x 10-5 m2/s
Water: γ m = 9780 N/m2
SOLUTION
Viscous correction is required when Rer < 500. To keep Rer > 500,
Rer = Ur/ ν
U > 500 ν /r = (500)(2 x 10-5 m2/s)/r
or we need
U > (0.01/r) m/s
From the pitot-static probe,
U=
2γ m H
ρair
Equating with the above expression for velocity (with r and H in [m]) gives,
H > (ρ / 2 γ m )(500ν / r) 2
> [1.225kg / m 2 /(2 × 9780N / m3 )] × [(500 × 2x10−5 m 2 / s) / r]2
For a 6 mm diameter probe (r = 0.003 m), U > 3.3 m/s (10.8 ft/s) and H > 0.7 mm.
PROBLEM 9.29
KNOWN: Anemometer circuit
R=
R=
500Ω ; α = 0.00395/oC
3
4
Sensor resistance at temperature Ts: Rs(20oC) = 110 Ω
FIND:
RD if Ts = 60oC
SOLUTION
For a metallic resistance temperature device, the resistance temperature relation is
approximated by,
Rs(Ts) = Ro[1 + α (Ts – To)]
Using Ro = R(20oC) = 110 Ω
Rs(60oC) = 110 Ω [1 + 0.00395(60 – 20C)]
= 127.38 Ω
If the sensor resistance is Rs, the bridge will be balanced when
Rs = RD(R3/R4) so that
RD = 127.38 Ω .
PROBLEM 9.30
KNOWN: Constant resistance anemometer
FIND:
Sensitivity relative to velocity
SOLUTION
For a thermal anemometer operating at constant resistance,
E2 = A + BUn
The sensitivity, K, is found by
K = (dE/dU)U = nBUn-1/2(A+BUn)1/2
For n = 0.5,
K ∝ U-0.75
The sensitivity decreases as velocity increases.
PROBLEM 9.31
KNOWN: F = 600 mm
θ = 5.5o
λ = 514.5 nm
FIND:
fd at U = 1, 10, 100 m/s
SOLUTION
fd = U [2 sin θ /2]/ λ = U [2 sin 2.75o]/514.5 x 10-9m = 186,540 U
At
U = 1 m/s, fd = 186.540 kHz
U = 10 m/s, fd= 1.8654 MHz
U = 100 m/s, fd = 18.654 MHz
Recomputing for F = 300 mm and θ = 7.3o,
fd = 247,517 U
At
U = 1 m/s, fd = 247.517 kHz
U = 10 m/s, fd= 2.47517 MHz
U = 100 m/s, fd = 24.7517 MHz
PROBLEM 9.32
KNOWN:
FIND:
LDA measurement using dual beam mode.
N = 5000
U = 21.37 m/s
θ = 6o
SU = 0.43 m/s
λ = 623.8 ± 0.5% nm
Bfd/fd ≤ 0.9%
B ≤ ≤ 0.25o
Estimate U', the best estimate of the velocity, based on available
information.
SOLUTION
The relation between velocity and Doppler frequency using dual beam mode is,
U = fd λ /[2 sin θ /2]
with the mean Doppler frequency estimated by
fd = U [2 sin θ /2]/ λ = (21.37 m/s)[2 sin 3o]/(623.8 x 10-9 m)
= 3.5858 MHz
The systematic error in the velocity estimate can be estimated by
 ∂U
± 
BU =
B
 ∂f d fd

Using
1/ 2
2
2
2
  ∂U
  ∂U
 
Bλ  + 
Bθ 
 + 
  ∂θ
 
  ∂λ

B θ = 0.25o = 0.0044 rad
Bfd = 0.009 fd = 32,272 Hz
B λ = (623.8 x 10-9 m)(0.005) = 3.12 x 10-9 m
1/ 2
2
2
2

fd
fd λ
 
 
λ
 
± 
BU =
Bf  + 
Bλ  + 
Bθ  
 2sin θ / 2 d   2sin θ / 2   sin θ / 2 tan θ / 2  
1/ 2
2
2
2
 623 × 10−9 m
  1.09 × 106 / s
  1.09 × 106 / s × 623.8 × 10−9 m
 
−9
9836 / s  + 
3.12 × 10 m  + 
0.0044  
=
± 
o
o
sin θ / 2 tan θ / 2
 2sin 3
  2sin 3
 
 
1/ 2
2
2
2
=
± ( 0.19 ) + ( 0.11) + ( 3.27 ) 


=
±3.28 m/s
The systematic error is completely dominated by the optical system error.
The random error in the mean velocity is estimated by
PU = SU/N1/2 = (0.43 m/s)/50001/2 = 0.007 m/s
with degrees of freedom of 4999.
The uncertainty in velocity is estimated by
uU = ± [B2 + (t4999,95PU)2]1/2
(95%)
= ± [3.282 + (1.96 x .007)2]1/2= 3.28 m/s
(95%)
The best estimate of the velocity may be stated as
U' = 21.37 ± 3.28 m/s
(95%)
COMMENT
The error in the optical set-up dominates the overall uncertainty in measuring velocity. One
can minimize this error by using a shorter focal length lens, as this will increase the value of
θ . But the value assigned for the systematic error in θ in this problem is actually quite
large for precise work. Standardized metrology equipment and methods allow vendors of
high quality optics to measure θ to better than 0.01o.
At B θ = 0.01o, U' = 21.37 ± 0.26 m/s (95%), a 1% uncertainty.
PROBLEM 9.33
KNOWN:
A = 4 m2 (2m by 2m duct)
pitot static tube measurements at 9 locations, each at the center of equal
rectangular areas
Air: T = 15 oC, p = 1 atm abs.
FIND:
Flow rate
SOLUTION
The flow rate Q = UA where the velocity is determined from each pitot-static tube
measurement. The dynamic pressure is measured by the pitot-static tube and it is related
directly to the velocity at the point of measurement by
pv = pt − p = ρ H 2O gH =
U=
1
ρ airU 2
2
2 ρ H 20 gH
ρ air
The value of H is obtained from the pitot-static tube. To estimate the average velocity in the
duct, a spatial average of the readings is made,
U=
2 ρ H 20 gH
ρ air
=
4
N
N =9
∑ H (mmH O) = 3.29 m/s
i =1
Then,
Q = (3.29 m/s)(4 m2) = 13.1 m3/s
2
PROBLEM 9.34
KNOWN:
Information of Problem 9.33
A = wL
Bw = 0.010 m
FIND:
BL = 0.010 m
Uncertainty in flow rate
SOLUTION
The flow rate is found from the continuity relation
Q = UA
so Q = UA = 13.1 m3/s (Problem 9.33) and uQ = f(uU, uA).
The instrument error associated with a pitot-static tube is assigned an uncertainty of 1 %,
BH/H = 1%
(Section 9.8)
The data variation in the 9 measurements of Problem 9.33 gives a random uncertainty in H
1/ 2
 9
2 
 ∑ (Hi − H ) 

PH =  i =1
9 −1




ν =8
/ 9 = 0.324 mm H2O
The uncertainty in flow rate contains systematic and random uncertainties. These propagate
as
1/ 2
2
2
 ∂Q
2 1/ 2
2
  ∂Q  


AB
+
UB
BQ 
BU  + 
B=
=
(
U )
A
A 



  ∂A  
 ∂U
(
1/ 2
)
 ∂Q  2  ∂Q  2 
2 1/ 2
2


AP
+
UP
PQ 
PU  + 
P=
=
(
U )
A
A 



 ∂U   ∂A  
But U =
2 ρ H 20 gH
ρ air
1/ 2
)
. If we consider only the errors in H as affecting the uncertainty in U,
2
 ∂U
 
 (4.0)(0.5) H −1/ 2 P
PU =
PH  
=

H


∂
H
 

1/ 2
(
(
)
2
 ∂U
 
 (4.0)(0.5) H −1/ 2 B
BU =
BH  
=

H


H
∂
 

(
2 1/ 2


)
= 0.263 m/s
2 1/ 2


=  2 H 1/ 2 BH / H  =0.222 m/s
A = wL:
1/ 2
1/ 2
 ∂A  2  ∂A  2 
( LB )2 + ( wB )2  = 0.028 m2
BA  
Bw  + 
B=
=

L 
w
L


 ∂w   ∂L  
PA = 0
Then,
1/ 2
(4 × 0.222) 2 + ( 3.29 × 0.028 )2 
BQ =


1/ 2
(4 0.263) 2 + ( 0 )2 
PQ =×


( ) 
uQ =
±  BQ2 + tPQ

2 1/ 2
= 0.893 m3/s
ν =8
= 1.026 m3/s
1/ 2
2
= 0.8932 + ( 2.306 × 1.026 ) 


Q = 13.1 ± 2.53 m3/s (95%)
There is about a 19% uncertainty in flow rate.
= ± 2.53 m3/s
(95%)
PROBLEM 9.35
KNOWN:
∆p ≈ 10 kPa
ρ Hg = S ρ H 2O where SHg = 13.6
FIND:
H if θ = 30o for inclined manometer
SOLUTION
Let H be the deflection in the vertical direction (such as a U-tube manometer),
10,000 N / m 2
∆p
∆p
∆p
=
= 0.075 m
≈ =
H=
γ m − γ γ m S Hg ρ H 2O g 1000kg / m3 9.8m / s 2 13.6
(
)(
)
But for an inclined manometer
H = L sin θ
or
L=
∆p
= 0.15 m
′ ρ H 2O g sin θ
S Hg
So the inclined manometer doubles the deflection for an applied head.
PROBLEM 9.36
KNOWN:
steel diaphragm transducer
t = 0.5 mm
d= 25 mm (so r = 12.5 mm)
ν p = 0.32
Em = 200 GPa
differential pressure limit ( ∆ p associated with ymax)
FIND:
SOLUTION
The differential pressure limit occurs at the maximum deflection of the diaphragm (after
which plastic deformation of the diaphragm may occur),
ymax =
3(∆p )(1 − ν 2p )r 4
16 Em t 3
where ymax = t/3. Then, rearranging
16 Em t 4
∆p =
9(1 − ν 2p )r 4
=
16(200 × 109 N / m 2 )(0.0005m)4
9(1 − 0.322 )(0.0125m)4
= 1 MPa
PROBLEM 10.1
KNOWN:
Flow of air through a pipe
U(r) = 25[1 - (r/r1)2] cm/s
p = 1 bar abs = 100,000 N/m2 abs
T = 5oC = 278 K
d1 = 2r1 = 5 cm
FIND:
mass flow rate
ASSUMPTIONS: Steady, incompressible, axisymmetric flow of a perfect gas.
SOLUTION
Conservation of mass gives
∂
ρd ∀ + ∫∫ ρ U • n̂dA = 0
∂t ∫∫∫
CV
CS
which for steady, incompressible, axisymmetric flow becomes
•
m=
2π
r1
∫ ∫ ρU(r)rdrdθ
0
0
The velocity can be written in vector form as
→
U = 25(1 −
r 2
) ez
r1
resulting in
•
m =ρ
2π
r1
r
∫ ∫ 25(1 − ( r )
0
0
2
) r dr dθ =
1
25
ρπr
2
2
1
For a perfect gas, the density can be estimated by
ρ = p/RT = 1.16 kg/m3
so that (with 1 m = 100 cm):
•
m = 4.5 × 10 −5 kg/s = 0.16 kg/hr
= 39.3ρ
cm 3 /s
PROBLEM 10.2
KNOWN:
Air flow through a pipe
diameter = 2r1 = 20 cm
N = 5 velocity measurements per cross-sectional traverse
M = 3 cross-sectional traverses
BQ/Q = ±2%
FIND: Q'
ASSUMPTIONS: Steady, incompressible flow of a perfect gas.
Flow rate remains perfectly controlled (constant) during all
measurements (equivalent to the assumption that all
measurements are taken simultaneously).
SOLUTION
The flow rate along each traverse line (j = 1, 2, 3) can be approximated by
r1
5
0
i =1
Q j = 2π ∫ Urdr ≈ 2π ∑ U ij r ∆r
i = 1,2,3,4,5
where ∆r = 2 cm and j = 1, 2, 3. Then,
Q=
2π [ (25.31)(1)(2) + (22.48)(3)(2) + (21.66)(5)(2) + (15.24)(7)(2) + (5.12)(9)(2) ]
1
= 4446 cm3/s
Similarly, Q2 = 4421 cm3/s, Q3 = 4400 cm3/s. The mean flow rate is
Q = (1/3)[4446 + 4421 + 4400]cm3/s = 4423 cm3/s
with standard deviation,
1/ 2
1 3

=
SQ  ∑ (Q j − Q ) 2 
 2 j =1

3
= 23 cm /s
and
SQ = SQ / 31/ 2 = 13.3 cm3/s
with ν = 2
Then, t2,95 = 4.303. With P = SQ , tP=57.2 cm3/s and with B = 0.02 Q = 88.5 cm3/s, then
uQ = [88.52 + 57.22]1/2 = 105.3 cm3/s.
So,
Q' = 4423 ± 105.3 cm3/s
(95%)
COMMENT
Sources of systematic error include: instrument errors, errors in the control of operating
conditions for all three traverses, and the approximation for the integral expression at the
start of the solution.
PROBLEM 10.3
KNOWN:
Manometer measuring the pressure drop of flowing water
H = 10.16 cm Hg
d1 = 5.1 cm
γ = 9800 N/m3 (water)
Sm = 13.57 (mercury) so that γ m = S mγ
FIND: p1 – p2
ASSUMPTIONS: p1 and p2 taps are located along the same horizontal datum line.
SOLUTION
Applying the hydrostatic equation between points 1 and 2 yields
p1 + L γ + H γ
p1 – p2 = H( γ
m
m
- (L+H) γ = p2
-γ)
= H(13.57 γ - γ ) = H γ 12.57
= (0.1016 m)(12.57)(9800 N/m2)(1 Pa/N/m2) = 12.516 kPa
PROBLEM 10.4
KNOWN:
Air flow through orifice meter
p1 – p2 = 69 kPa
d1 = 25.4 cm
T = 32 oC
FIND: H
SOLUTION
p1 – p2= γ H
H = (p1 – p2)/ γ
with γ = 9750 N/m3 (Appendix)
H = (69,000 N/m2)/9750 N/m3 = 7.041 m H2O = 704.1 cm H2O
PROBLEM 10.5
KNOWN: Water flow through orifice meter using flange taps.
d1 = 3 in.
T = 60oF
p1 = 100 psi
p2 = 76 psi
do = 1.5 in
RO2 = 48.3 ft-lb/lbm-oR
FIND: Is Y < 1?
ASSUMPTIONS: Steady flow
SOLUTION
Y = f(k, β , ∆p / p1 )
For a diatomic gas, such as O2, k = 1.4.
β = d o / d1 = 0.5
∆p / p1 = 0.24
Using Figure 10.7,
Y = 0.92
There is an 8% reduction in flow rate due to compressibility effects.
COMMENT
Aside from the usual sources of error in an measurement, data reduction errors enter into the
obstruction meter relations from the assumed values of the various coefficients and from the
ability to read these values from the tables and charts. Normally, these are treated as
systematic errors unless additional information about how they were estimated is known.
PROBLEM 10.6
KNOWN:
Orifice meter using flange taps
do = 5 cm
d1= 15 cm
Red1 = 250,000
FIND: C
SOLUTION
C = f( β , Red1)
β = d o / d1 = 0.33
Red1 = 250,000
From Figure 10.6, Ko = CE = 0.60.
E = 1/(1 - β 4 )1/2 = 1.006
C = Ko/E = 0.596 ~ 0.60
COMMENT
Aside from the usual sources of error in a measurement, data reduction errors enter into the
obstruction meter relations from the assumed values of the various coefficients and from the
ability to read these values from the tables and charts. Normally, these are treated as
systematic errors unless additional information about how they were estimated is known.
PROBLEM 10.7
KNOWN: Flow of water through orifice meter at 20oC
d1 = 10 cm
β = 0.4
FIND: Q at which C = f( β , Red1) becomes C = f( β )
ASSUMPTIONS: Flange pressure taps are used so that Figure 10.6 is applicable.
SOLUTION
For β = 0.4, Figure 10.6 shows Reynolds number independence in flow coefficient for
Red1 > 20,000. Because Ko = CE and because E depends only on β , we conclude that C =
f( β ) for all Red1 > 20,000.
Re d 1 = 4Q / πν d1 > 20,000
1
Q > πν d1Re =
π (1 x 10-6 m2/s)(0.1m) (20000)/4 = 1.6 x 10-4 m3/s
d
1
4
with ν from the Appendix.
COMMENT
Aside from the usual sources of error in an measurement, data reduction errors enter into the
obstruction meter relations from the assumed values of the various coefficients and from the
ability to read these values from the tables and charts. Normally, these are treated as
systematic errors unless additional information about how they were estimated is known.
PROBLEM 10.8
KNOWN: Water flow through an orifice plate with flange taps
Q = 50 L/s
T = 25oC
d1 = 12 cm
β = 0.5
FIND: ∆p, ∆ploss
ASSUMPTIONS: Steady, incompressible (Y = 1) flow.
SOLUTION
For an orifice meter,
=
Q CEAY 2∆p / ρ with A based on do. Now, do = β d1 = 0.06m and
A = π d o2 / 4 = .0028m2:
Ko = CE = f( β , Red1) = f(0.5, 530,000) ~ 0.63
Using ν = 1 x 10-6 m2/s and ρ = 997 kg/m3 from the Appendix ,
Re d 1 = 4Q / πν d1 = 4 (0.05m3)/ ( π x 1 x 10-6 m2/s)(0.12m) = 530000
Then,
∆p =
( ρ / 2 )( Q / CEAY ) = [(1000 kg/m3)/2]{(0.05m3/s)/(0.63)(0.0028m2)(1)}2
= 4.017 x 105 N/m2
2
From Figure 10.8, we can expect
∆ploss = 0.77∆p = 3.093 x 105 N/m2
COMMENT
Aside from the usual sources of error in an measurement, data reduction
errors enter into the obstruction meter relations from the assumed values of the
various coefficients and from the ability to read these values from the tables and
charts. Normally, these are treated as systematic errors unless additional information about
how they were estimated is known.
PROBLEM 10.9
KNOWN:
Air flow through a flow nozzle (k = 1.4)
do = 3 cm
d1 = 6 cm
H = 75 cm H2O
p1 = 94.4 kPa
T1 = 38 oC
FIND: Q
ASSUMPTIONS: Steady flow of a perfect gas.
SOLUTION
For a flow nozzle,
Q CEAY 2∆p / ρ with A based on do. Now, β =do/d1 = 0.5 and A =
=
π d o2 / 4 = .00071m2.. With λ = 9800 N/m3 from the Appendix,
∆p =
γ H = (9800 N/m3)(0.75 m) = 7350 Pa
∆p / p1 = 7350 Pa/94900 Pa = 0.078
The compressibility factor (Figure) is Y = f(k, β , ∆p / p1 ) ~ 0.98. For a perfect gas,
ρ1 = p1 / RT1 = (94,400 Pa)/(287 J/kgK)(311) = 1.06 kg/m3
We need a value for Ko. But Ko = f( β , Red1) and Red1 depends on Q.
Guess a value for Ko and iterate on a solution. From Figure 10.11, pick a value of Ko that lies
on the β = 0.5 curve. Lets choose Ko = 1.008, which is a value that lies in the flat region of
the curve. Then,
Q = (1.008)(0.00071m2)(0.98)[(2)(7350 Pa)/(1.06kg/m3)]1/2 = 0.082 m3/s
But this uses a guessed value. So, check on the guessed value of Ko. Using our new value for
Q:
Re d 1 = 4Q / πν d1 = (4)(0.078 cms)/( π )(0.06m)(1.6x10-5 m2/s) = 103,000
or from Figure 10.11, K~ 1.008. Close enough.
So the solution remains: Q = 0.082 m3/s.
COMMENT
Aside from the usual sources of error in a measurement, data reduction errors enter into the
obstruction meter relations from the assumed values of the various coefficients and from the
ability to read these values from the tables and charts. Normally, these are treated as
systematic errors unless additional information about how they were estimated is known.
PROBLEM 10.10
KNOWN: Nitrogen (k = 1.4) flows through an orifice meter with flange taps.
T1 = 520oR
p1= 20 psia
p2 = 15 psia
d1 = 4 in.
β = 0.5
RN2 = 55.13 ft-lb/lbm-oR
FIND: Q
ASSUMPTIONS: Steady flow of a perfect gas
SOLUTION
For an orifice meter,
Q CEAY 2∆p / ρ with A based on do. Now, do = β d1 = 2 in. and A
=
= π d o2 / 4 = .0218ft2. So find Y, E, and C:
Y(k, (p1 – p2)/p1, β )
p1 – p2 = (20 – 15) psia = 5 psia = 720 psf
so,
Y(k, (p1 – p2)/p1, β ) = Y(1.4,0.25,0.5) = 0.92
and
(p1 – p2)/p1 = 0.25
(Fig. 10.7)
ρ1 = p1 / RT1 = (20 psi)(144 in2/ft2)/(55.13 ft-lb/lbm-oR)(520oR) = 0.1 lbm/ft3
Now, Ko = f( β , Red1) and Red1 depends on Q. Guess a value for Ko and iterate on a solution.
From Figure 10.16, pick a value of Ko that lies on the β = 0.5 curve. Lets choose Ko = 0.63
Q = (0.63)(0.0218 ft2)(0.92)[(2)(32.2 lbm-ft/lb-s2)(5 lb/in2)(144 in2/ft2)/0.1lbm/ft3]1/2
= 8.6 cfs
Checking: Re d 1 = 4Q / πν d1 = (4) (8.6 ft3/s)/ ( π )(1.15 x 10-4 ft2/s)(4/12 ft) = 285655
and Ko (285,655 and 0.5) ~ 0.63.
So, Q = 8.6 ft3/s.
PROBLEM 10.11
KNOWN: Water flow through an orifice meter
T = 20oC
d1 = 38 cm
m ≈ 200 kg/s
H ≤ 15 cm Hg ( with S = 13.57)
FIND: Design orifice meter by setting do.
ASSUMPTIONS: Steady, incompressible (Y = 1) flow
SOLUTION
The mass flow rate for a constant density ( ρ ) flow
=
m ρ=
Q ρ CEAY 2∆p / ρ
with A based on orifice diameter, do. Solving for do with A = π d o2 / 4 ,
1/ 2
d o  4m /(CEπρ 2∆p / ρ 
=
From the Appendix , ρ = 999 kg/m3 and µ = 9.6 x 10-4 N-s/m2.
=
∆
p γ Hg=
H S ρ gH
= (13.57)(9760 N/m3)(0.15m) = 19928 N/m2
If we were to select β = 0.5, then do = 19 cm (trial). Then
Re d 1 = 4m / πµ d1 = (4)(200 kg/s)/ π (0.38m)(9.6x10-4N-s/m2) = 698000
From Figure 10.6, Ko = f( β , Red1) = f(0.5, 698000) ~ 0.625, then,
d o ≥ [(4)(200kg/s)/{(.625) ( π )(1000kg/m3)[(2)(19928N/m2/(10000kg/m3)]1/2}]1/2
= 0.25 m or 25 cm
(updated value)
Using this new value for do gives β = 0.658.
With β = 0.658, do = 25 cm. Rechecking: Ko(0.658, 698000) ~ 0.625 and do = 0.25 m.
Converged. So, this value for do will meet our constraints.
We may wish to build in a safety factor on the ∆ p constraint. So setting β = 0.7 or do = 26.6
cm (Note from Figure 10.6 that this is the largest value for β at which the ASME tables may
be used. Larger β values will require in situ calibration. We find Ko = 0.668 and
the resulting pressure head works out to 10.8 cm Hg at 200 kg/s.
COMMENT
It is apparent that by choosing different values for do, the pressure drop across an obstruction
meter can be altered as desired for an expected flow rate.
PROBLEM 10.12
KNOWN: Water flowing through a venturi meter
T = 15oC
d1 = 10-cm.
H ≤ 76 cm. H2O
Q ≈ 0.5 m3/min = 0.0083 cms
FIND: Design the meter size, do
ASSUMPTIONS: Steady, incompressible (Y = 1) flow.
Cast venturi meter.
SOLUTION
From (10.14), Q = CEAY(2∆p/ρ1)1/2 where A and β (in the term E) are based on the
venturi throat diameter, do. Using ρ = 998 kg/m3 and ν = 1x10-6 m2/s for water from
Appendix C:
∆p = γH2O H = (ρg/gc)H ≤ (998 kg/m3)(9.8 m/s2) (76 cm/100 cm/m) (1 N/kg-m/s2) = 7433
N/m2
We seek,
do ≥ [4Q/{CEπ (2∆p /ρ1)1/2}]1/2
Using Red1 = 4Q/πd1ν = 4(0.0083cms)/ π(10/100)(1.0x10-6 m2/s) = 106000. We choose β=
0.5 so that for the venturi, C = 0.984 (see text). Then, E = 1/(1 - β4 )1/2 = 1.0328. So,
do ≥ [4(0.0083 cms)/{(.984)(1.0328) π [2 (7433 N/m2 )/(998kg/m3)]1/2 }]1/2 = 0.052m or 5.2
cm
This yields a β sufficiently close to 0.5, as selected. A value of do ≥ 5.2 cm will meet the
constraint.
COMMENT
It should be apparent that by choosing different values for do the pressure drop across an
obstruction meter can be altered. Hence, one can specify the dimensions based on an
expected flow rate so as to achieve a desired range of values. 16
PROBLEM 10.13
KNOWN: Q = 120 cfm of water at 60oF
H ≤ 20 in. Hg
d1 = 6 in.
pump efficiency: η = 0.60
power costs: $ 0.10/kW-hr
operating time: 6000 hr/yr
FIND: Specify design for suitable orifice, venturi and nozzle.
∆ploss and operating costs for each device.
ASSUMPTIONS Steady, incompressible (Y= 1) flow.
PROPERTIES water: ρ1 = 62.4 lbm/ft3 ν = 1 x 10-5 ft2/s
SOLUTION
For an obstruction flow meter, Q = CEAY(2∆p/ρ1)1/2where ∆p =ρ gH = 1411 lb/ft2.
At Q = 120 cfm,
Re d 1 = 4Q / πν d1 = (4)(120cfm)(1min/60s)/( π )(.5ft)(1x10-5 ft2/s) = 5.1x105
The meter throat sizes do are found from
do = [(4Q)/(CE π 2 ∆p /ρ1 ) ]1/2
Orifice:
From Figure 10.6, Ko = f( β , Red1). If we guess, Ko = 0.6, then do = 0.33 ft for
β = do/d1 = 0.66. From Figure 10.6, Ko(0.66, 5 x 105) = 0.65, so that do = 0.32 ft for β =
0.64. Then from Figure 9.6, Ko(0.64, 5 x 105) = 0.65. So β = 0.64
and do = 3.84 in.
From Figure 10.8, with β = 0.64 and Red1 = 5 x 105,
∆ploss = 0.6 ∆p = 847 psf
ASME Long Radius Nozzle:
From Figure 10.11, Ko = f( β , Red1). If we guess, Ko = 1.0, then do = 0.26 ft for
β = do/d1 = 0.52. From Figure 10.11, Ko(0.52, 5 x 105) = 1.01, so that do = 0.26 ft. So β =
0.52 and do = 3.12 in.
From Figure 10.8, with β = 0.52 and Red1 = 5 x 105,
∆ploss = 0.6 ∆p = 847 psf (Same as the orifice, but with smaller β )
Venturi:
Suppose we choose a 15o cast model. The text states, based on the ASME Power Test Codes,
that C ~ 0.98.
So we begin an iteration with a guess, Ko = CE = 0.98. Then, do = 0.265 for a β = 0.53.
E = (1- β 4)-1/2 = 1.04. So we get Ko = 1.02. Using Ko = 1.02, do = 0.26 ft, β = 0.52 and
E = 1.04. This is close enough. So, do = 3.12 in.
From Figure 10.8,
∆ploss = 0.18 ∆p = 254 psf
Cost Analysis
The power use is estimated from:
W = Q∆ploss /η
and operating cost from:
cost = (W)(operating time)(cost/kW-unit time)
β
orifice
nozzle
venturi
0.64
0.52
0.52
W [kW]
3.7
3.7
0.9
Cost[$/yr]
2230
2230
538
PROBLEM 10.14
KNOWN:
Water flow at 27oC through an ASME long radius nozzle.
β = 0.6
H = 25.4 cm Hg
d1 = 15 cm.
FIND: Q
ASSUMPTIONS: Steady, incompressible (Y=1) flow.
PROPERTIES: Water: ρ = 997 kg/m3 ν = 1 x 10-6 m2 /s
SOLUTION
From (10.14) with A and β based on throat diameter do,
Q = CEAY(2∆p/ρ1)1/2
With β = do/d1 = 0.6, do = (0.6)(15) cm. = 9 cm. = 0.09 m and
∆p = γH2O H = (ρg/gc)H = (997 kg/m3)(9.8 m/s2) (0.09m) (1 N/kg-m/s2) = 879 N/m2
The value for Ko = CE = f(β, Red1). From Figure 10.11, guess Ko = 1.035. Then, with
A = π(0.09)2/4 = 0.0064 m2 and Y =1 (incompressible water):
Q = (1.035)(0.0064 m2) (1)[2(879 N/m2 )(1 kg-m/s2-N) /997 kg/m3 )]1/2
= 0.0088 cms
Check the guess:
Red1 = 4(0.0088cms)/π(.09m)(1x10-6 m2/s) = 124,500
From Figure 10.11, Ko = (0.6, 124,500) = 1.03.
So, Q = 0.0088 cms
PROBLEM 10.15
KNOWN: Orifice meter
d1 = 9 cm
do = 4 cm
Meter to be located within 4 m of straight run of pipe.
Upstream elbow.
FIND: Meter placement
ASSUMPTIONS: Elbow is in-plane with downstream pipe.
SOLUTION
For β = do/d1 = 0.44, we use Figure 10.13 which indicates that the meter
should be located in a straight run of pipe 6.25 diameters downstream of the
elbow with 3 diameters of straight run pipe downstream of the meter. Using 7
diameters upstream to be conservative, this should just fit into the available
space.
PROBLEM 10.16
KNOWN: Orifice meter used as a sonic nozzle to meter air flow.
T1 = 40oC
p1 = 695 mm Hg abs = 92,660 N/m2 abs
p2 = 330 mm Hg abs = 43,996 N/m2 abs
d1 = 5 cm = 0.05 m
β = 0.4
FIND:
mass flow rate
PROPERTIES: R = 287 N-m/kg K ; k = 1.4
ASSUMPTIONS: Air behaves as a perfect gas.
SOLUTION
For this flow,
p2/p1 = 330 mm Hg/695 mm Hg = 0.475
(p2/p1)crit = [2/(k+1)]k/(k-1) = 0.528
Since p2/p1 < 0.528, the flow in the orifice throat is choked and sonic conditions exist there.
Using,
ρ1 = p1/RT1 = (92,660 N/m2 abs)/( 287 N-m/kg K)(300 K) = 1.076 kg/m3
A = πdo2/4 = π(0.4 x 0.05)2/4 = 0.0003 m2
Equation 10.18 applies using k = 1.4, R = 287 N-m/kg K, and T1 = 300 K:
•
m = ρ 1 A(2RT1 ) 1/2 [(k/k + 1)(2/k + 1) 2/k −1 ]1/2 = 0.067 kg/s
PROBLEM 10.17
KNOWN: Air flow through an orifice meter with flange taps.
do = 0.5m
d1 = 1m
H = 90 mm H2O
p1 = 2 atm = 3 atm absolute
FIND: Q
ASSUMPTIONS: Steady flow of a perfect gas.
PROPERTIES: air: ν = 1.5 x 10-4 m2/s R = 0.287 kJ/kg-K
water: ρ = 1000 kg/m3
SOLUTION
Using equation 10.14 with A and β based on the orifice throat diameter do,
Q = CEAY(2∆p/ρ1)1/2
where ρ1 = p1 / RT1 = (3)(101,325 N-m2/atm)/(287 J/kg-K)(293K) = 3.6 kg/m3
∆p =ρ gH = (1000 kg/m3)(9.8 m/s2)(0.09 m)/1 kg-m/N-s2 = 882 N/m2
With β = do/d1 = 0.5 and ∆ p/p = 0.003, Figure 10.7 gives Y ~ 1. From Figure 10.6, guess
Ko = f( β ,Red1) = 0.633. Then,
Q = (0.633)( π (0.5 m)2/4)[2(887 N/m2)/1000 kg/m3]1/2 = 2.75 m3/s
Checking, Red1 = 4Q/ π d1ν = 4(2.75 m3/s)/ π (1m)(1.5x10-4 m2/s) = 23,350
From Figure 10.6, Ko(0.5, 23350) ~ 0.635 or Q = 2.76 m3/s.
Checking, Red1 = 23,400 and Ko = 0.635.
So, Q = 2.76 m3/s.
PROBLEM 10.18
KNOWN:
Flow of 20oC water through an ASME Long Radius Nozzle.
β = 0.5
d1 = 8 cm id
0.6 ≤ Q ≤ 1.6 liters/s (note: 1 liter = 0.001 m3)
Transducer:
u∆p /∆p: 0.25% full scale (design-stage uncertainty)
FIND: ∆p or H; design stage uncertainty in Q
ASSUMPTIONS: Steady, incompressible (Y = 1) flow of water.
PROPERTIES: Water: ρ = 997 kg/m3; ν = 1.0 x 10-6 m2/s
SOLUTION
Rewriting equation 10.14,
∆p = (ρ1/2)(Q/CEYA)2
Using do = βd1 = (0.5)(8) = 4 cm and A = πdo2/4 = π (.04)2 /4 = 0.00126 m2
At 0.6 liters/s,
9550
Red1 = 4Q/πd1ν = 4(0.6 liters3/s)(0.001m3/liter) /π (.08)(1.0 x 10-6 m2/s) =
Ko(0.5, 9550) = CE = 0.965
Then, ∆p = (997 kg/m3/2)[(0.0006 m3/s)/(0.965)(0.00126)]2 = 121 N/m2 = 0.9 mm Hg
At 1.6 liters/s, Red1 = 4Q/πd1ν = 4(1.6 liters3/s)(0.001m3/liter) /π (.08)(1.0 x 10-6 m2/s) =
25,450
Ko(0.5, 25450) = CE = 0.965
Then, ∆p = (997 kg/m3/2)[(0.0016 m3/s)/(0.965)(0.00126)]2 = 863 N/m2 = 6.5 mm Hg
So a transducer with a range from about 0 to 7 mm Hg (0 to 96 mm H2O) is needed.
Electronic pressure transducers are readily available with a range from 0 to 10 mm Hg, and
such a transducer will be suitable. (Alternatively if a mechanical output is acceptable, a
manometer filled with water would work well in this range at a fraction of the cost.)
The propagation of uncertainty to the flow rate can be written as,
u Q = ±[(
∂Q
∂Q
∂Q
∂Q
∂Q
uC ) 2 + (
uE )2 + (
uA )2 + (
uρ )2 + (
u ∆p ) 2 ]1 / 2
∂∆p
∂C
∂E
∂A
∂ρ
or dividing by Q = CEA(2∆p/ρ)1/2, the uncertainty on a percent basis is
uQ
Q
= ±(
uC
C
)2 + (
uE
E
)2 + (
uA
A
)2 + (
uρ
2ρ
)2 + (
u ∆p
2∆p
) 2 ]1 / 2
For the design-stage uncertainty analysis, we assume:
Transducer:
Discharge coefficient:
Diameters:
Density:
u∆p /∆p = 0.0025
uC/C = 0.02
ud1/d1 = udo/do = 0.002
uρ /ρ = 0.002
[given]
[text ref. 2]
[reasonable assumption]
[reasonable assumption]
Then,
Area:
uA/A = 2udo/do = 0.004
Beta ratio:
uβ /β = [(udo/do)2 + (ud1/d1)2]1/2 = .003
Approach factor: uE/E = [2β3/(1-β4)2]uβ /β = 0.0009
uQ/Q = ± [0.022 + 0.00092 + 0.0042 + (0.002/2)2 + (0.0025/2)2]1/2 = ± 0.02
That's an uncertainty of about 2% of the flow rate.
COMMENT
Note how the known information about the instruments and the coefficients has been used to
estimate this uncertainty. It is clear that the uncertainty in discharge coefficient dominates the
uncertainty in flow rate. In situ calibration of the nozzle could reduce this uncertainty in C.
The uncertainty due to the pressure transducer is too small to be an important factor in this
measurement.
PROBLEM 10.19
KNOWN: Water at 60oF flowing through an orifice meter.
10 ≤ Q ≤ 50 gpm
d1 = 2.3 inch
u∆p / ∆p =± 0.005
Select C = f( β ) only
FIND: do; design stage uncertainty in Q
ASSUMPTIONS: Steady, incompressible (Y = 1) flow of water.
PROPERTIES: Water: ρ = 62.4 lbm/ft3
ν = 1.2 x 10-5 ft2/s
SOLUTION
Begin by finding do.
For an orifice meter, Q = CEA(2∆p/ρ)1/2=
. E 1/ 1 − β 4 . From Figure 10.6, K becomes
essentially independent of Reynolds number at the higher values. So we will try to meet the
constraint at the lowest expected flow rate so that it will be met at all higher flow rates.
At 10 gpm,
Red1 = 4Q/πd1ν= 4(10 gpm)(448 cfs/gpm)/ π (1.2x10-5 ft2/s)(2.3/12 ft) = 9900
Again, using Figure 10.6 at Re = 9900, select β = 0.3. The value for discharge coefficient
falls within a range (C = Ko/E): 0.60 ≤ C ≤ 0.615. Then with do = 0.6 inches, we can
assume a constant value of C = 0.607 to within 2.5% over the flow range.
The uncertainty of a result, in this case the flow rate, is estimated by propagation of terms
and uncertainties, Q = f(C, E, A, ρ , ∆ p). This gives
uQ
Q
= ±(
uC
C
)2 + (
uE
E
)2 + (
uA
A
)2 + (
uρ
2ρ
)2 + (
u ∆p
2∆p
) 2 ]1 / 2
(equation 1)
From Chapter 10 reference 2 or 12, we find that we can estimate the error in discharge
coefficient at ± 6% of C, i.e., B1 = ± 0.006C, when using the tabulated values. We also
assume that C is constant over the full flow range, which introduces a systematic error
estimated at B2 = 2.5% of C. Together we obtain the systematic uncertainty in C,
1/ 2
( B1 / C ) 2 + ( B2 / C ) 2 
BC / C =
1/ 2
(0.006) 2 + (0.025) 2 
=
= 0.026
Further, it is reasonable to expect a tolerance of 0.005 inch in all dimensional
measurements. Hence, we use this value as an estimate of the systematic error in the
dimensional terms,
Bdo/do = 0.0083
Bd1/dd1 = 0.0025
Then, expanding each uncertainty and plugging in values gives
=
π d o Bdo / 4 2 Bdo / d o = 0.0167
BA / A 2=
1/ 2
Bβ / β ( Bdo / d o ) 2 + ( Bd 1 / d1 ) 2  = 0.017
=
2
BE / E  2β 3 /(1 − β 4 ( Bβ / β )  = 0.0009
=


(
)
It is reasonable to assume a ± 0.5% systematic uncertainty in density, so
Bρ / ρ = 0.005
We will treat all of the uncertainty in the pressure transducer as a systematic uncertainty, then
B∆p / ∆p = 0.005 ( because we have no other information to treat it otherwise). Subbing back
into equation (1),
1/ 2
BQ / Q =
± 0.0262 + 0.0167 2 + 0.00092 + (0.005 / 2) 2 + (0.005 / 2) 2 
=
± 0.031
Finally, recognizing that all errors are treated as systematic errors, such that all P = 0, then
uQ / Q = BQ / Q = ± 0.031
The uncertainty in flow rate is about 3.1%.
COMMENT
Certainly, the uncertainty in C dominates the uncertainty in Q. It is worth
noting that as Red1 approaches 50 000 and beyond (Q > 40 gpm), the value for
C approaches 0.60. Using this value for all flow rates would retain the previous
uncertainty value at low flow rates but would reduce uQ/Q to less
than 1.8% at the higher flow rates. Try it!
We treat all errors in this problem as systematic errors and assign them systematic
uncertainties. We then propagate these systematic uncertainties through to a result. Had we
simply assigned the values as uncertainties (i.e., replace all the B terms with u terms), the
problem would develop exactly the same. When errors can be separated into random and
systematic components, their separation serves a useful purpose.
PROBLEM 10.20
KNOWN: Air (70oF) flow through an orifice
10 ≤ Φ ≤ 80% where Φ is relative humidity
ρ (Φ = 45%) used
FIND: uQ due to variation in density due to humidity changes
PROPERTIES: Using a psychiometric chart for air at standard atmosphere,
ρ (80%) = 0.0746 lbm/ft3
ρ (10%) = 0.0735 lbm/ft3
ρ (45%) = 0.0741 lbm/ft3
SOLUTION
Inspection of the properties reveals that the density can be estimated as
ρ = 0.0741 ± 0.0005 lbm/ft3
over the range of relative humidity values. This does not account for systematic errors
in the actual density values obtained from the chart. For an orifice plate,
Q = CEA(2∆p/ρ)1/2
To isolate the effects of humidity on the flow rate, we neglect all error sources except those
involved with density (as density is affected by humidity). This assumes that C, A, and
pressure are not affected by humidity variation, which is reasonable. Then,
1/ 2
 ∂Q 2 
uQ = f ( ρ ) = ± 
uρ  
 ∂ρ  
Dividing through by the flow rate, gives
(
)
1/ 2
2
uQ / Q = ±  uρ / 2 ρ 


= 0.0034
or an uncertainty of about 0.34% of the flow rate results due to relative humidity effects
alone.
PROBLEM 10.21
KNOWN: Air flow through an orifice meter with the conditions from Problem
10.20.
Q = 17 m3/hr = 17 cmh
T = 20oC
β = 0.4
d1 = 6 cm
u∆p / ∆p = 0.005
udo = ud1 = 0.1 mm
p1 = 96.5 kPa
FIND: Design stage uncertainty in flow rate
PROPERTIES R = 28 N-m/kg-K
ν = 1.6 x 10-5 m2/s
k = 1.4
SOLUTION
For an orifice, Q = CEA(2∆p/ρ)1/2 with do = β d1 = 2.4 cm. But Ko= CE = f( β , Re). At 17
cmh, Red1 = 4Q/ πν d1 = 6260. Then from Figure 10.6, Ko(0.4, 6260) = 0.63 (at least to
within ± 0.005). To solve equation (10.14) for ∆ p, we assume Y = 1. This yields for A =
0.00045 m2,
∆ p = [0.0047 cms/(.63)(0.00045m2)(1)]2(1.15 kg/m3/2) = 157.85 N/m2
Using this value, ∆ p/p1 = 0.0016. With k = 1.4, Figure 10.7 gives Y = 1. So ∆ p = 157.85
N/m2. This is equivalent to about 1.6 cm H2O.
Based on (10.14) and the propagation of uncertainty to a resultant, dividing by (10.14), the
uncertainty in flow rate can be written as
uQ
u
u
u
u
u
u
=
±( C ) 2 + ( E ) 2 + ( A ) 2 + ( Y ) 2 + ( ρ ) 2 + ( ∆p ) 2 ]1/ 2
2ρ
2∆p
Q
C
E
A
Y
The uncertainty in C is due both to the intrinsic error in using tabulated values,
u1, and due to our ability to read Figure 10.6, u2. Setting u1 = 0.006C (see text)
and u2 = 0.008C (based on an estimated resolution of ± 0.005 from the chart):
uC/C = (0.0062 + 0.0082)1/2 = 0.01
Other values:
udo/do = 0.01/2.4 = 0.004
ud1/d1 = 0.01/6 = 0.0017
uA/A = 2udo/do = 0.008
u / β = [ (udo/do)2 + (ud1/d1)]1/2 = 0.0043
β
uE/E = 2 β 3(1 - β 4)-2u / β = 0.0009
β
uY/Y = 0.004 ∆ p/p1 = 1.3 x 10-5
(see text and reference 2)
u ∆p / ∆p = 0.005
Lastly, the uncertainty in density will be due both to the error from the presumed humidity
variations, u1, and the error in the tabulated values, u2. Taking u1/ ρ = (0.0005/1.15) = 0.0004
and u2/ ρ = 0.005 (i.e. about 0.5% of the tabulated value):
u ρ / ρ = [0.00042 + 0.0052]1/2 = 0.005
Then,
1/ 2
uQ / Q =
± 0.012 + 0.00092 + 0.0082 + (1.3 × 10−5 ) 2 + (0.005 / 2) 2 + (0.005 / 2) 2 
= ± 0.0133
Accounting for known uncertainties, the uncertainty is about 1.3% of the flow rate.
COMMENT
The assignment of uncertainty estimates requires time, some research into available
information and good common sense. Keep in mind that the above estimate does not include
the effects of control of operating conditions and the measurement procedure.
PROBLEM 10.22
KNOWN: Air flow at 70oF through an ASME Long Radius Nozzle
Q = 45 cfm
p1 = 14.1 psia
FIND: p2 and do required to assure choked flow at the throat.
ASSUMPTIONS: Air behaves as a perfect gas in this process
PROPERTIES: R = 53.3 ft-lb/lbm-oR
k = 1.4
SOLUTION
From (10.16), the critical pressure ratio is given by
[p2/p1]c= (2/k+1)k/k-1 = (2/2.4)1.4/0.4 = 0.528
The critical pressure ratio sets the largest pressure possible and still choke the
throat. Then, the critical downstream pressure is
[p2]c = 0.528p1 = 7.45 psia
So, p2 ≤ 7.45 psia to choke the flow.
Under choked flow conditions, the mass flow rate is at a maximum value. Equation 10.18
gives the mass flow rate at the critical pressure ratio which must then be the maximum mass
flow through the nozzle. Rearranging (10.18), we solve for the throat area at critical
conditions which must represent the largest throat area that can choke the flow,
Amax =
m / ρ
2 2 / k −1
k
2 RT
(
)
k +1 k +1
(45 ft 3 / min× 1min/ s )
=
2 × 53.3
1.4
2 2 /(1.4 −1)
ft − lb
530o R × 32.2lbm − ft / s 2 − lb
(
)
o
1.4 + 1 1.4 + 1
lbm R
= 0.0011 ft2
But A = π d o2 / 4 . So do ≤ 0.459 inches would provide a suitable design.
PROBLEM 10.23
KNOWN: Water at 20oC flows through orifice plate with flange taps.
d1 = 10 cm
H = 100 mm Hg at Q = 2 m3/min = 0.033 cms
FIND: do required
SOLUTION
We want to have H = 100 mm Hg or ∆ p = (13.57)(9800N/m2)(0.1 m) = 13,300 N/m2.
Equation 10.14 can be rewritten as:
d o = [4Q/(CEπ (2 ∆p/ρ1 ) 1/2 )] 1/2
for the conditions known, this reduces to do= 0.09/(Ko)1/2 [m]
To solve, we that Ko = CE = f( β , Red1). For the target flowrate,
Red1 = 4Q/ πν d1=4(0.033 cms)/ π (.1m)(1x10-6 m2/s) = 425,000
Inspection of Figure 10.6 shows that the orifice diameter requires a β of nearly 1 to satisfy
the flow meter equation. So an orifice meter can not be used to meet the given constraints! If
an orifice meter were used, the pressure drop would be substantially less than the desired
value.
PROBLEM 10.24
KNOWN: Nozzle used to meter 20oC water.
d1 = 20 cm
5000 cm3/s ≤ Q ≤ 50,000 cm3/s
β = 0.5
FIND: Pressure transducer range required.
SOLUTION
Rearranging the obstruction meter equation (10.14) with Y = 1 (incompressible):
( ρ / 2)(Q / K o A) 2
∆p =
where Ko = f( β =0.5, Red1).
Low Q: Red1 = 4(.005 m3/s)/ π (.2m)(1x10-6 m2/s) ~ 32,000
High Q: Red1 = 4(.05 m3/s)/ π (.2m)(1x10-6 m2/s) ~ 320,000
So from Figure 10.11: (Ko)low ~ 1.000 ; (Ko)high ~ 1.005.
Solving:
∆plow = 203 N/m2
∆phigh = 20,100 N/m2
So the selected transducer must have a range extending from about 200 to 20,500 N/m2, one
rated from 0 to 21 cm H2O will work well.
PROBLEM 10.25
SOLUTION
The solution to this problem is open-ended. We suggest that the instructor assign this
problem and use it as a basis for an in-class or small group discussion.
PROBLEM 10.26
KNOWN: Vortex meter metering fluid flow
St = 0.20 (shedder Strouhal number)
f = 77 Hz (shed frequency)
d = 1.27 cm (shedder characteristic length, see Table 10.1)
FIND: Average velocity, U
SOLUTION
St = fd/U
U = (77 Hz)(0.0127)/0.2 = 4.89 m/s
PROBLEM 10.27
KNOWN: Air flow at 30oC metered by a thermal mass flow meter
d1 = 2 cm
Power = P = 25 W
∆ T = 1oC = 1K
cp = 1.006 kJ/kg-K
FIND: mass flow rate
ASSUMPTIONS: cp is constant through meter and known.
∆ T reported (measured) must be the average mixed temperature across each pipe
cross section.
Power supplied to meter is 100% dissipated in fluid (i.e. no losses).
SOLUTION
 p ∆T
=
E mc
But the energy supplied to the meter is dissipated into the flow, so that E = P .
Then,
=
m P / c p ∆T = (25 W)(1006 J/kg-K)(1K) = 0.025 kg/s
COMMENT
Note the many assumptions that go into using this meter. Actually, for flows of gases which
can be well modeled as perfect gases, these types of meters provide excellent accuracy. For
other types of fluids, they have limited applicability.
PROBLEM 10.28
KNOWN:
Sonic nozzle used to regulate volume flow rate of air.
FIND:
Uncertainty due to normal changes in ambient temperature and pressure.
SOLUTION
From (10.18),
•
m = ρ 1 A(2RT1 ) 1/2 [(k/k + 1)(2/k + 1) 2/k −1 ]1/2
For air, (p2/p1) cr = 0.528. Provided that the actual p2/p1 ratio remains below this value the
throat will remain choked. From (10.3) and (10.6), we can rewrite this for volume flow rate
as,
Q = A(2RT1)1/2 [(k/k+1)(2/k+1)2/k-1]1/2
This shows that the volume flow rate will be affected by environmental changes in
temperature even though constant mass flow rate is maintained in the nozzle throat (all other
terms in the equation remain constant). If we look only at the effects of temperature variation
over the ± 5 K range specified, then Q = f(T) only. This can expressed as
uQ/Q = ±[(uT/2T)2]1/2
= ± [(5 K)/(2 x 283 K))2 ]1/2
= ±0.0088
or an added uncertainty in volume flow rate of 0.88% or about 1% due to temperature
changes alone.
PROBLEM 10.29
KNOWN: From Example 10.4: Q = 0.053 cms (air)
d1 = 6 cm
β = 0.4
do = 2.4 cm
H = 250 cm H2O
p1 = 93.7 kPa
T=1 = 293 K
Y = 0.92
All dimensions known to ± 0.025 mm
Upstream pressure is constant
Pressure drop has systematic uncertainty B = ± 0.25 cm H2O
SH = 0.5 cm H2O with N = 20
FIND: uQ
PROPERTIES: water: γ = 9800 N/m3; ρ = 998 kg/m3
air: ν = 1.6 x 10-5 m2/s; ρ = 1.16 kg/m3
SOLUTION
The flow rate is found by
Q = f(C,E,A,Y, ∆ p, ρ )
with Q = CEA(2∆p/ρ)1/2 and ∆ p = (9800 N/m3)(2.5 m H2O) = 24.5 kPa
The systematic uncertainty in Q can be estimated on a percent basis as,
BQ
B
B
B
B
B
B
=
±( C ) 2 + ( E ) 2 + ( A ) 2 + ( Y ) 2 + ( ρ ) 2 + ( ∆p ) 2 ]1/ 2
2ρ
2∆p
Q
C
E
A
Y
where
Bdo/do = 0.0025/2.4 = 0.001
Bd1/dd1 = 0.0025/6 = 0.0004
B β =[(Bdo/do)2 + (Bd1/d1)2]1/2 = 0.001
BA/A = 2Bdo/ddo = 0.002
BE/E = 2 β 3(1 - β 4)-2B β / β = 0.0002
BY/Y = 0.004 ∆ p/p1 = 1.3 x 10-5 (see text and reference 2)
B ∆ p/ ∆ p = 0.25/250 = 0.001
B ρ / ρ = 0.005
(assumed value)
According to the text and reference 2, the systematic uncertainty in using the tabulated values
for C amounts to 0.006C. However, we will also add an additional elemental systematic error
in reading Figure 10.6 estimate with systematic uncertainty of 0.008C. Then,
BC/C = (0.0062 + .0082)1/2 = 0.0094
Inserting yields,
BQ/Q = [0.00942+0.00022+0.0022+0.0000132+(0.005/2)2+(0.001/2)2]1/2 = 0.01
Or BQ = 0.00053 m3/s. Note how the systematic error in C dominates.
Likewise the random uncertainty in Q can be expressed as
PQ
P
P
P
P
P
P
=
±( C ) 2 + ( E ) 2 + ( A ) 2 + ( Y ) 2 + ( ρ ) 2 + ( ∆p ) 2 ]1/ 2
2ρ
2∆p
Q
C
E
A
Y
But the only random error estimate comes from the pressure reading, so that the random
uncertainty is
PQ/Q = P∆p / ∆p
With SH = 0.5 cm H2O ,
PH = SH/(20)1/2 = 0.11 cm H2O
or in correct units,
P∆p = 9.8 N/m2 with a degrees of freedom of 19.
Then,
(
)
1/ 2
2
= [(9.8/(2)(24500))2]1/2 = 0.0002
PQ/Q =  P∆p / 2∆p 


Or PQ = 1x10-5 m3/s.
Combining random and systematic uncertainty estimates for flow rate,
1/ 2
uQ =± 0.00532 + (2.093 × 1 × 10−5 ) 2 
= ± 0.00053 cms (95%)
COMMENT
The uncertainty amounts to about 1% of the flow rate and is primarily due to systematic
errors in the variables, notably the discharge coefficient. The user may feel more comfortable
using larger values for the systematic errors as read from charts and properties obtained from
tables. The point is that a careful consideration of each term relevant to the measurement has
been made.
PROBLEM 10.30
KNOWN: Q = 10W
∆T =
3o C
Air: cp = 1.006 kJ/kg K
FIND: m
ASSUMPTION: Air is a perfect gas (i.e., cp = constant)
SOLUTION
=
m
Q
10W
=
(1J / s / W ) = 0.0033 kg/s = 0.2 kg/min
c p ∆T (1006 J / kgK )(3o C )
PROBLEM 10.31
KNOWN: St = 0.19 (Table 10.1 for Red > 10,000)
d = 10 mm = 0.010 m
D = 10 cm = 0.10 m
U = 30 m/s
T = 20oC air
FIND: K1, Q
SOLUTION
=
f
St × U 0.19 × 30m / s
= 5,700 Hz
=
d
0.01m
=
K1 π=
d / 4St π (0.01m) /(4 × 0.19) =0.0413
Q = K1D2f = (0.0413)(0.1m)2(5700Hz) = 0.236 m3/s
Check: Q = UA = U π d 2 / 4 = 0.236 m3/s
ok
Check: Re = ρUd / µ = 20,000 so St = 0.19 ok
PROBLEM 10.32
KNOWN:
Q = 30 acmm = 30 actual m3/min
p = 50 mm Hg
T = 15oC
FIND:
Qstandard
SOLUTION
ρ
ρ s tan dard
actual
Qs tan dard Q=
Qactual (
=
actual
293
760 + 50
)(
)
273 + 15C
760
Qs tan dard = 1.0843Qactual = 32.5 scmm
PROBLEM 10.33
KNOWN:
Cast venturi
C = 0.984
do = 3.995 in
d1 = 6.011 in
ρ = 62.369 lbm/ft3
H = 100 in H2O
with
with
with
with
with
b = 0.00375
b = 0.0005
b = 0.001
b = 0.002
b = 0.15
P=0
P=0
P=0
P = 0.002 (large dataset)
P = 0.4 (large dataset)
FIND: Best estimate of the flow rate at 95% confidence
ASSUMPTION: Neglect thermal expansion of the venturi material; Dataset is large such
that N > 30, so that t ~ 2 is used throughout the problem
SOLUTION
This problem is taken from PTC 19.1 – 2005 where it was revised, edited and presented in
that standard by author R. Figliola. The problem is formatted in a manner recommended by
the standard, although the nomenclature has been revised to be consistent with this text.
 =
m
ρQ =
ρCEYA 2∆p / ρ
where ρ is the density of water. With ∆p =ρgh and h is the equivalent head in terms of
inches of water; Y is set equal to 1 for incompressible liquids; E = 1/ 1 − (
d 4
) ; and
D
A = πd 2 / 4 , we can express the mass flow rate in lbm/s as
0.099702Cd 2 ρh
 =
m
4
d
1−  
D
 f (C, d, ρ, h, D) and u = f (B , B , B , B , B ; P , P , P , P , P ) , or using the fact
So,
=
m

C
d
ρ
h
D C d ρ h D
m
that b = B/2 (i.e., the standard systematic uncertainty equals one-half the systematic
uncertainty) we can equivalently write u  = f (bC , bd , bρ , b h , b D ; PC , Pd , Pρ , Ph , PD ) . Also,
m
1/ 2
2
2
2
2
2
 




  ∂m
  ∂m
  ∂m
  ∂m
 
∂
m
b m = 
b  +
b  +
b  +
b  +
b  
 ∂C C   ∂d d   ∂ρ ρ   ∂h h   ∂D D  


1/ 2
2
2
2
2
2
 
   ∂m
   ∂m
   ∂m

 
∂m   ∂m

Pm = 
P  +
P  +
P  +
P  +
P  
 ∂C C   ∂d d   ∂ρ ρ   ∂h h   ∂D D  


so that the uncertainty in mass flow rate is found by
1/ 2
1/ 2
2
2
(95%)
u m =
± ( B ) + (2P) 2  =
± ( 2b ) + (tP) 2 




where t has been assigned the value of t ~ 2. Or, in terms of the standard uncertainty
1/ 2
2
u m =
± ( b ) + (P) 2 


(68%)
Table 1 states the known information. Table 2 states the sensitivity terms and values used in
the propagation formulas. Table 3 states the contributions of each variable to the uncertainty
in mass flow rate. Table 4 states the results.
We find (from Table 4) that
 138.4 ± 1.21 lbm/s
=
m
(95%)
which is equivalent to the result
 138.4 ± 0.608 lbm/s
=
m
(68%)
Either answer is a correct statement at its specified probability level.
Table 1: Nominal Values and Uncertainties for each Variable
Independent Parameters
X
Symbol
C
d
D
ρ
h
Description
Discharge
Coefficient
Throat Diameter
Inlet Diameter
Water density @
60oF [25]
Differential
pressure head
across venture
(68oF)
Units
Nominal
Value
Absolute
Systematic
Standard
Uncertainty
bX
Random
Standard
Uncertainty
of the Mean
PX = s
X

0.984
3.75E-03
0
inches
inches
lbm/ft3
3.999
6.001
62.37
5.0E-04
1.0E-03
0.002
0
0
0.002
in.H2O
100
0.15
0.4
GENERAL NOTES:
a) The systematic and random estimates for density are based on water temperature
measurements having systematic and random uncertainties of 0.2oF and 0.1oF, respectively.
b) The systematic uncertainty for the differential pressure head is assumed to be one-half
the least count of the manometer scale.
Table 2: Estimation of the Sensitivities for Each Variable
Sensitivity

∂m
∂X
Formulas for Absolute Sensitivity
X
Nominal
Values
C
0.984

∂m
∂X
0.099702d 2 ρh
140
4
d
3.999
D
6.001
d
1−  
D

3 
d




0.099702C ρh d 2 ( −4 )   
4

d
D 
1 −   (2)0.099702C ρhd − 
4
4

D
d  d  
 2 1 −   D 1 −    

 D    D   

(
−
)
4(0.099702)Cd 6 ρh
  d 4 
2 D 1 −   
 D 


72.6
-11.4
3/ 2
5
ρ
h
62.37
0.099702Cd
2
4
ρ
100
0.099702Cd
h
2
d
1−  
D
1.11
2
d
1−  
D
h
ρ
2
4
0.692
Table 3: Uncertainty Values for Each Variable
Independent Parameters
Parameter Information
(in Parameter Units)
X
Description
C
Discharge
coefficient
d
D
ρ
h
Units
Nominal
Value
Absolute
Systematic
Standard
Uncertainty
Absolute
Random
Standard
Uncertainty
Absolute
Sensitivity
bX
PX = Sx
θX
Uncertainty Contribution
of
Parameters to the Result
(in Result Units Squared)
Absolute
Absolute
Systematic
Random
Standard
Standard
Uncertainty
Uncertainty
Contribution
Contribution
(b X θX ) 2
(PX θX ) 2

0.984
3.75E-03
0
140
0.276
0
Throat
diameter
inches
3.999
5.0E-04
0
86.2
1.86×10-3
0
Inlet
diameter
inches
6.001
1.0E-03
0
-11.4
1.29×10-4
0
Water
density @
60oF [25]
lbm/ft3
62.37
0.002
0.002
1.11
4.93×10-6
4.92×10-6
Differential
pressure
head across
venturi (at
68oF)
in.H2O
100
0.15
0.4
0.692
1.08×10-2
7.66×10-2
Table 4: Uncertainty Values: Results
Description

m
Mass Flow
rate
Units
lbm/s
Calculated
Value
138.4
Absolute
Systematic
Standard
Uncertainty
Absolute
Random
Standard
Uncertainty
b m
Pm
0.537
0.276
Combined
Standard
Uncertainty of
the Result
Total Absolute
Uncertainty
u m (68%)
u m (95%)
0.604
1.21
COMMENT
The tables here are set up in a convenient manner to express the formulation and results of
the problem. This format meets that recommended by ASME/ANSI PTC 19.1.
PROBLEM 10.34
KNOWN:
Flow nozzle
d1 = 60 mm = 0.060m
Q = 0.003 m3/s
ρ = 790 kg/m3
µ = 1.2 x 10-3 N-s/m
FIND:
do
ASSUMPTION: Incompressible flow (Y = 1)
SOLUTION
=
Q CEAY 2∆p / ρ = K o (π d o2 / 4) 2∆p / ρ
0.003m3 / s = K
π
4
d o2
2(4000 N / m 2 )
790kg / m3
rearranging
d o2 =
1.20 × 10−3
Ko
We know Ko = CE = f(Red1, β )
ρ Q / π d1µ
=
Re 4=
4 × 789kg / m3 × 0.003m3 / s
= 42,200
π × 0.06m × 1.2 × 10−3 N − s / m
From Figure 10.11, with Re = 42,200, then 0.98 ≤ K o ≤ 1.02
Guess Ko = 1.00, then
d0 = 1.2 × 10−3 / K o = 0.0346 m
Then, β = 0.0346/0.06 = 0.58
From Fig 10.11, Ko = f(42,300, 0.58) ~ 1.01
Then, do = 0.0345 m; β = 0.057.
From Fig 10.11, Ko = f(42,300, 0.57) ~ 1.01
So, do = 0.0345 m = 34.5 mm
PROBLEM 11.1
KNOWN: A steel rod (circular cross-section) having:
L = 10 in
D = 0.25 in
Em= 30 ×106 lb/in 2
m = 40 lb m
FIND: The change in length of the rod, δ L ,
ASSUMPTIONS: Rod is elastically deformed, such that
σ = ε ⋅ Em
SOLUTION:
The force resulting from 40 lbm in standard gravity is
ma
F
=
=
N
gc
( 40 lbm ) ( 32.174 ft/sec2 )
ft lb m 

 32.174

lb sec 2 

= 40 lb
The resulting uniaxial stress is
σa =
FN
Ac
D=0.25 in
10 in
where
π
=
( 0.25) 0.049 in 2
4
40 lb
σa =
=
814.9 lb/in 2
2
0.049 in
=
Ac
2
Cross
Section, Ac
and
=
εa
σa
814.9 lb/in 2
2.716 ×10−5
=
=
6
2
Em 30 ×10 lb/in
The change in length is then
δ L =L ⋅ ε a =( 2.716 ×10−5 ) (10 in ) =0.00027 in
FN
PROBLEM 11.2
KNOWN: A steel rod (circular cross-section) having:
L = 0.3 m
D = 5 mm
E=
20 ×1010 Pa
m
m = 50 kg
FIND: The change in length of the rod, δ L .
ASSUMPTIONS: Rod is elastically deformed, such that
σ = ε ⋅ Em
SOLUTION:
The force resulting from 50 kg in standard gravity is
ma
F
=
=
N
gc
( 50 kg ) ( 9.8 m/s 2 )
kg m 

1.0 2 
s N

FN = 490 N
D=5 mm
0.3 m
The resulting uniaxial stress is
σa =
Cross
Section, Ac
FN
Ac
FN
where
π
1.96 ×10−5 m 2
Ac = ( 5 ×10−3 ) =
4
490 N
σ=
25 ×106 Pa
=
a
-5
2
1.96 ×10 m
2
and
ε=
a
σa
25 ×106 Pa
=
= 125 ×10−6
10
Em 20 ×10 Pa
The change in length is then
δ L =L ⋅ ε a =( 0.3) (125 ×10−6 ) =37.5 ×10−6 m
PROBLEM 11.3
KNOWN: An electrical coil with
N = 20, 000
D = 0.051 in
r = 2.0 in
FIND: The resistance, R.
SOLUTION:
We know
ρe L
R=
Ac
where ρ e = 1.673 ×10−6 Ω cm for copper, and
=
Ac
π
4
( D=
)
2
π  0.051 
2
−5
2
=

 1.42 ×10 ft
4  12 
The length is then found as
 2
L 2=
000 ) 20,944 ft
=
=
π r ( N ) 2π   ( 20,
 12 
which yields a resistance of
(1.673 ×10
R=
= 81 Ω
R
−6
Ω-cm ) ( 0.0328 ft/cm )( 20,944 ft )
1.42 ×10−5 ft 2
PROBLEM 11.4
KNOWN: Aluminum having a volume of 3.14159 x 10-5 m3, and a resistivity,
ρ e = 2.66 ×10−8 Ω m .
FIND: Resistance, R, of 2 mm and 1 mm diameter wires having the same total volume.
SOLUTION:
Volume for a cylindrical wire is
V=
π
(D ) L
4
2
yielding
L2mm = 10 m and L1mm = 40 m
The resistance values are then calculated as
=
R2 mm
ρe L
π
=
Ac
D2
Ac
4
m)
( 2.66 ×10 Ω-m ) (10=
π
( 2 ×10 m )
4
m)
( 2.66 ×10 Ω-m ) ( 40=
π
(1×10 m )
4
−8
R2 mm
=
−3
2
−3
2
0.085 Ω
−8
R1mm
=
1.355 Ω
PROBLEM 11.5
KNOWN: A nickel conductor with
ρe= 6.8 x 10-8 Ω m
Ac = 5 x 2 mm (rectangular)
L=5m
FIND:
a) R - the total resistance
b) The diameter of a 5 m long copper wire having a circular cross-section
to yield the same resistance.
SOLUTION:
The resistance is found from
=
R
ρe L
2
=
=
Ac 10 mm
L 5m
Ac
which in this case yields
=
R
( 6.8 ×10
Ω m) (5 m)
= 0.034 Ω
10 ×10−6 m 2
−8
For the copper, ρ e =1.7 ×10−8 Ω m
0.034 =
(1.7 ×10
−8
Ω m) (5 m)
Ac
−6
Ac =
2.5 ×10 m
2
D=
1.8 mm
PROBLEM 11.6
KNOWN: A Wheatstone bridge with all resistances initially equal to 100 Ω. The
maximum power through R1 is 0.25 W.
FIND:
Maximum applied voltage
Bridge sensitivity
ASSUMPTIONS: Infinite meter resistance
SOLUTION: From the circuit shown below
i1 R1 + i2 R2 =
Ei
i1= i2= i
i ( R1 + R2 ) =
Ei
But we know power, P, is given by P = i 2 R , and
=
i
0.25 W
= 0.05 A
100 Ω
At node A
ii = i1 + i3 = 2i ii = 0.1 A
Ei = ii RB where RB is the equivalent bridge resistance
=
so Ei ( 0.1 A=
)(100 Ω ) 10 V
The bridge sensitivity is defined as
δ E0
KB =
δ R1
and for a bridge with all resistances initially equal and assuming δ R  R
δ E0 Ei R
=
≈
δ R 4R
(10 V )(100 Ω ) ≈ 2.5 V/Ω
400 Ω
PROBLEM 11.7
KNOWN: A strain gauge with R1 = 120 Ω, GF = 2 in an equal arm Wheatstone bridge
R2 = R3 = R4 = 120 Ω. Maximum gauge current is 0.05 A
FIND: Maximum input bridge voltage
SOLUTION:
From a basic circuit analysis, assuming infinite meter resistance
i1 =
Ei
R1 + R2
and
=
Ei i1 ( R1 + R2 )
=
Ei
A )( 240 Ω )
( 0.05 =
12 V
PROBLEM 11.8
KNOWN: A strain gauge has a nominal resistance of 350 Ω, and GF = 1.8, and senses
axial strain. The gauge is mounted on a 1 cm2 aluminum rod (Em = 70 GPa) Eo = 1 mV,
Ei = 5 V.
FIND: Applied load, assuming uniaxial tension
SOLUTION:
For an equal arm bridge, from (11.14)
(δ R R )
(δ R R )
δ E0
0.001
=
=
4 + 2 (δ R R )
5
4 + 2 (δ R R )
Ei
and
δ R=
R 0.0008
=
δ R 0.28 Ω
Since
δR R=
ε ⋅ GF
ε=
0.00044
and with
σ
ε=
σ =ε ⋅ Em =( 0.00044 )( 70 GPa )
Em
σ = 0.0308 GPa
To find the applied force, FN
2
FN
π
σ=
A c =×
1 10−2 m ) =
7.854 × 10−5 m 2
(
4
Ac
FN =
( 7.854 ×10−5 m2 )( 0.0308 ×109 Pa )
and since 1 Pa = 1N m 2
FN = 2419 N
COMMENT:
This force would result from a mass, in standard gravitational
acceleration, of 246.6 kg.
PROBLEM 11.9
KNOWN: Strain gauge installation shown in Figure 11.12.
FIND: Show that this installation is not sensitive to bending stresses.
SOLUTION:
The stresses created by a bending and axial load may be represented as
Compression
+
Tension
Tension
In general, for four active elements in a bridge,
δ E0
=
Ei
GF
( ε1 − ε 2 + ε 4 − ε 3 )
4
or for this case
δ E0 GF
=
( ε1 + ε 4 )
Ei
4
Strain may be expressed as a linear combination of the imposed loads
δ E0
=
Ei
(
GF 
ε m + ε FN
4 
) + (ε
1
m
+ ε FN
) 
4
But since ε m1 = −ε m4 this installation is not sensitive to bending.
PROBLEM 11.10
KNOWN: A steel member ( ν p = 0.3) subject to simple axial tension. Strain gauges
are mounted on top center, and bottom center. GF
= 2, All R=
' s 120 Ω
δ=
Eo 10 µ V and=
Ei 10 V,=
R 120 Ω and GF =2
FIND: Bridge constant, for gauge locations 1 and 4. Is the system temperature
compensated? Determine the axial and transverse strains.
SOLUTION:
The configuration is
1
4
FN
FN
Since for any 4 gauges
δ E0
=
Ei
GF
( ε1 − ε 2 + ε 4 − ε 3 )
4
and for a single gauge, sensing maximum strain
δ E0
Ei
=
GF
( ε max )
4
In the present case both gauges sense the maximum strain, and the outputs are additive
κ =2
This installation is not temperature compensated.
For the gauge sensing maximum strain
δ E0 s
Ei
 ε max GF 
=

 4 + 2ε max GF 
The actual output is then
δ E0 = κδ E0 s
and
δ E0 s = 5 V
Solving for ε max yields
ε max = 1×10−6
Therefore
ε axial = ε max = 1×10−6
ε=
0.3 ×10−6
t
PROBLEM 11.11
KNOWN: An axial and a transverse strain gauge are mounted to the top surface of a
steel beam, and connected in arms 1 and 2 of a Wheatstone bridge.
=
δ E0 250 µ V
=
υ p 0.3
=
σ 2222.2 psi
=
Ei 10 V =
Em 29.4 × 106 psi
FIND: a) κ B
b) Average gauge factor
SOLUTION:
In general
δ E0
GF
( ε1 − ε 2 + ε 4 − ε 3 )
4
=
Ei
which in this case, for one axial and one transverse gauge yields
δ E0 GF
=
( ε max − 0.3ε max )
4
Ei
=
GF
( 0.7ε max )
4
for a single gauge sensing the maximum strain
δ E0
Ei
=
GF
ε max
4
and
κ B = 0.7
To find the average value of GF
δ E0
Ei
=
GF
ε maxκ B
4
250 ×10−6 GF
=
ε maxκ B
10
4
with
σ = 2222.2 psi and ε max = σ Em
=
ε max 7.559 ×10−5
and
4 ( 250 ×10−6 )
=
GF =
1.89
10 ( 0.7 ) ( 7.559 ×10−5 )
Comment: The chosen arrangement of strain gauges yields a bridge constant less than
one, which without other considerations, is not a good choice.
PROBLEM 11.12
KNOWN: A strain gauge, mounted on a steel cantilever, has the following characteristics:
=
R 120 Ω
R 0.1 Ω
δ=
GF
= 2.05 ± 1% (95%)
uR = ±1%
FIND: Estimate the strain, εa, and the uncertainty in the measured strain, uε.
ASSUMPTIONS: The bridge operates in a null mode and reasonable values for input
voltage and galvanometer sensitivity must be assigned.
SOLUTION:
For an equal arm bridge,
δ R1
= ε a ⋅ GF
R1
which yields
0.1
= ε=
ε a 0.000407
a ( 2.05 )
120
The uncertainty analysis can be approached in several ways. Since the bridge is operated
in a null mode, a galvanometer and a calibrated resistor are employed. The following
relationships are used for the bridge


δR R
= 

Ei
 4 + 2 (δ R R ) 
R1 = R2 ( R3 R4 ) (balanced bridge)
δ E0
Let
γ = ( δ R1 R1 )
A typical galvanometer sensitivity may be ±1 µV, and Ei = 10 V, then
ε=
γ
GF
12
2
 ∂ε 2  ∂ε
 
=
uE  uγ  + 
uGF  
 
 ∂γ   ∂GF
This equation for the uncertainty in strain contains two uncertainties, yet to be estimated.
The partial derivatives which represent the senstivity indices are evaluated at the nominal
values as
∂ε
=
∂γ
∂ε
∂GF
1
1
=
GF 2.05
γ
0.1
=
−
=
−
=
2 × 10−4
2
2
GF
( 2.05) (120 )
δ R1
0.1 Ω
=
R1 120 Ω
where=
γ
We must examine the uncertainty in γ, which has contributions from the galvanometer, and
from the bridge and calibrated resistors.
The analysis proceeds as
 4 ( δ E0 Ei ) 

1 + 2 ( δ E0 Ei ) 
γ =
uγ =
b
∂γ
uδ E
∂ ( δ E0 Ei ) i
E0
∂γ
At δ E0 0= 4
=
∂ ( δ E0 Ei )
Assume the only contribution to δ E0 E1 is the galvanometer, ± 1 µV ⇒
uδ E0
Ei
=
±1.1× 10−7 V
Additional contributions to uncertainty in γ result from
Bridge resistance R3, R4 ± 1%
Calibrated resistor R2 ± 1%
and with R1 = R2 ( R3 R4 )
∂R1 R3
= = 1
∂R2 R4
∂R1 R2
= = 1
∂R3 R4
RR
∂R1
=
− 2 23 =
1
∂R4
R4
3 ( 0.01) =
uR1 =
±0.0173 Ω
2
0.173
a
uγ =
=
±0.000144 Ω
120
a
b
Combining uγ and uγ
 4 (1× 10−7 )  + [ 0.000144]2 =
±0.000144 Ω
uγ =


2
Then with
12
2
2
 1
  γ
 
=
uε 
uγ  +  −
u
2 GF  
 
 GF   GF
12
2
2
 1
  0.1 120
 
0.000144  +  −
0.02  
= 
2
  2.05
 
 2.05
With this result, the uncertainty in strain is found as
uε =±7 ×10−5
PROBLEM 11.13
KNOWN:
R = 120 Ω
Gauges mounted on opposite arms of bridge
Ei= 4 V
Eo = 120 µV
GF = 2
Em = 29 x 106 psi
FIND: Resistance change for each gauge
SOLUTION:
For this bridge
δ E0 GF
=
ε1 − ε 2 + ε 4 − ε 3
Ei
4
which implies a bridge constant of 2.
(
)
Thus
δ E0 GF
κ GF
=
=
( 2ε a ) B ( ε max )
4
4
Ei
 120 ×10−6   4 
−5
=
 3.0 ×10
 
4

  2 ( 2) 
= 
ε max
Since
δR
R
= ε max GF
δ R=
(120 Ω ) ( 3 ×10−5 ) ( 2 )
The change in resistance is
=
δ R 0.0072 Ω
PROBLEM 11.14
KNOWN:
A rectangular bar in uniaxial tension, having strain gages mounted to measure axial and
transverse strain. Cross-section, Ac = 2 in2, axial strain, εa = 1500 µε, transverse strain,
εa = − 465 µε, axial load, FN = 1500 lb
FIND: Modulus of elasticity, Em, and Poisson’s ratio, νp
SOLUTION:
The axial stress and strain are related as
σ a = Emε a
1500 lb
750 lb/in 2
=
2 in 2
yielding
=
σa
Em =
σ a 750 lb/in 2
=
= 5 ×105 lb/in 2
ε a 1500 ×10−6
Poisson’s ratio is determined from the transverse strain,
=
νp
transverse strain
465
= = 0.31
axial strain
1500
PROBLEM 11.15
KNOWN:
A circular bar in uniaxial tension, having strain gages mounted to measure axial and
transverse strain. Cross-section, Ac = 3 cm2, axial strain, εa = 600 µε, transverse strain,
εa = − 163 µε, axial load, FN = 10 kN
FIND: Modulus of elasticity, Em, and Poisson’s ratio, νp
SOLUTION:
The axial stress and strain are related as
σ a = Emε a
10 kN
33.3 MPa
=
3 × 10-4 m 2
yielding
=
σa
E=
m
σ a 33.3 MPa
=
= 55.6 GPa
ε a 600 ×10−6
Poisson’s ratio is determined from the transverse strain,
=
νp
transverse strain 163
= = 0.27
axial strain
600
PROBLEM 11.16
KNOWN: A single active gauge and a dummy gauge are employed to measure strain.
The active gauge experiences an axial loading with bending, and both strain gauges
experience the same temperature.
FIND: Show that the use of the dummy gauge compensates for temperature, but not for
bending
SOLUTION:
In general
δ E0
=
Ei
GF
( ε1 − ε 2 + ε 4 − ε 3 )
4
The active gauge will experience axial, bending and temperature effects:
ε1 = ε a + ε b + ε T
The dummy gauge will experience temperature effects only
ε 2 = εT
Consider the case where the gauges are mounted on adjacent arms of a Wheatstone
bridge. Then,
δ E0
Ei
=
GF
GF
[ε1 − ε 2 ] = ε a + ε b + ε T − ε T = [ ε a − ε b ]
4
4
Axial strain is measured. Temperature effects are compensated, but bending effects
remain.
PROBLEM 11.17
KNOWN: Four active gauges mounted on a cylindrical shaft as per Table 11.1
FIND:
Show axial, temperature and bending compensation
SOLUTION
Each active gauge will experience torsional, axial, bending and temperature effects:
ε1 = εθ + εa + εb + εT
ε3 = -εθ + εa + εb + εT
ε2 = -εθ + εa + εb + εT
ε4 = εθ + εa + εb + εT
Then, when mounted on a wheatstone bridge (such as in Figure 11.23), are balanced, and
a load applied, the output voltage deflection is given by:
δ Eo
GF
[ε 1 − ε 2 + ε 4 − ε 3 ] = GFε θ
Ei
4
The arrangement measures torsional strain. The arrangement compensates for axial,
bending and temperature effects. Note that this arrangement has a bridge constant of 4.
=
PROBLEM 11.18
KNOWN: Bridge arrangements of Figure 11.23
FIND: Bridge constants
SOLUTION:
Using equation 11.22
δ E0
=
Ei
GF
( ε1 − ε 2 + ε 4 − ε 3 )
4
For a single gauge sensing the maximum strain
δ E0
Ei
a)
κB
=
b)
=
κB
=
GF
ε max
4
( GF 4 ) ε1 1
=
( GF 4 ) ε max
GF 4 )( ε1 − ε 3 ) ( GF 4 ) ε1 − ( −υ pε1 ) 
(=
( GF 4 ) ε1
( GF 4 ) ε max
κ B = 1+υ p
c)
κB
=
GF 4 )( ε1 − ε 3 ) ( GF 4 ) ε1 − ( −ε1 ) 
(=
=
( GF 4 ) ε1
( GF 4 ) ε1
2
d)
κB =
( GF 4 )( ε1 − ε 2 + ε 4 − ε 3 )
( GF 4 ) ε1
ε2 =
−υ pε1
ε3 =
−υ pε 4
κ=
2 (1 + υ p )
B
e)
κB =
( ε1 − ε 2 + ε 4 − ε 3 )
ε1
−ε1
−ε 4
ε2 =
ε3 =
κB = 4
PROBLEM 11.19
KNOWN: Four active gauges mounted to a diaphragm so as to measure displacement
R1 and R4 sense tension when R2 and R3 sense compression and gauges are
mounted as in Figure 11.23
GF = 2.0; R = 120Ω; ε = 20µs; Ei = 9V
δEo
FIND:
SOLUTION
Each active gauge will experience axial and temperature effects:
ε1 = εa + εT
ε3 = -εa + εT
ε2 = -εa + εT
ε4 = εa + εT
Then, when mounted on a wheatstone bridge, balanced, and a load applied:
δ Eo
Ei
=
GF
[ε 1 − ε
4
2
+ε
4
− ε 3 ] = GFε
Then,
δE
o
= (2)(20 x 10 6 )(9V) = 360 μV
PROBLEM 11.20
KNOWN: Conditions of Problem 11.20. Oops, R2 and R4 wiring is interchanged.
R1 and R2 sense tension when R4 and R3 sense compression and gauges are
mounted as in Figure 11.23
δEo
FIND:
SOLUTION
Each active gauge will experience axial and temperature effects:
ε1 = εa + εT
ε3 = -εa + εT
ε2 = εa + εT
ε4 = -εa + εT
Then, when mounted on a wheatstone bridge, balanced, and a load applied:
δ Eo
Ei
=
GF
[ε 1 − ε
4
2
+ε
4
−ε 3] = 0
The bridge deflection is zero. No load is sensed!
PROBLEM 11.21
KNOWN: D = 1 m, Transverse sensitivity = 0.03 = 3%
FIND: Error due to transverse sensitivity
SOLUTION:
Since
PD
εt σ l
PD
1
and σ=
then = = 4t=
l
4t
ε a σ t PD 2
2t
From Figure 11.6, the error can be determined as 2.8% of σt.
PD
σ=
t
2t
PROBLEM 11.22
KNOWN: Wheatstone bridge circuit, with all fixed resistances equal to 100 Ω.
R1 is a strain gauge with a resistance of 100 Ω at zero strain. The strain gauge senses ε1.
Maximum power dissipation in gauge = 0.25 W
=
t 2=
cm D 2 =
m GF 2
FIND:
Maximum allowable static sensitivity, in V/kPa
Under what conditions is K constant?
SOLUTION:
From problem 11.6, the maximum value of Ei to limit power dissipation to 0.25 W is
10 V, and the static sensitivity is 2.5 V/Ω, based on the resistance change of the strain
gauge. In the present problem, we can write
PD
σ
σl =
=
and ε l
Em
4t
PD
then ε l =
4 Emt
δR
and with
= ε GF , then
R
δ R PD
=
GF
R 4 Emt
Taking the derivative of δ R with respect to P yields
RD
GF
4 Emt
and the static sensitivity is then
RD
GFK B where KB = bridge sensitivity in V/Ω = 2.5
4 Emt
thus the static sensitivity for input pressure is
(100 Ω )( 2 m )
2 ( 2.5
=
mV/Ω ) 0.0000625 V/Ω or 0.0625 mV/Ω
( 4 ) ( 20 ×107 kPa ) ( 0.02 m )
PROBLEM 11.23
KNOWN: A Wheatstone bridge measurement system is to be designed to measure the
tangential strain in the wall of a pressure vessel. A reasonable estimate of the resulting
uncertainty is desired. The bridge is to be operated in a balanced condition.
FIND: The bridge input voltage, the fixed resistance values, and the galvanometer
sensitivity.
SOLUTION:
The following provides an outline for addressing this design problem. Assuming that
R=
R=
R=
R=
120 Ω at balanced conditions and zero strain,
1
2
3
4
The change in R1 with applied strain can be expressed
δR
= ε GF
R
The mathematical relations for a balanced bridge are
I g ( R1 + Rg )
u
R2 R4
and with the resistances equal, R =
=
R1
Ei
R1 R3
The input voltage must be designed based on a trade-off between static sensitivity from
the uncertainty, and the power that must be dissipated in the bridge. The strain gauge
must dissipate I12 R1 , and should serve as the limiting factor for the input voltage.
PROBLEM 11.24
KNOWN: Steel cantilever beam (fixed at one end) equipped with four active gauges
R1 and R4 mounted on top; R2 and R3 on bottom
Gauges attached to bridge circuit as per Figure 11.23
F = 980N; A = 1000; GF =2; L = 0.1m; b = 0.3m; t = 0.01m, Ei = 5V
δE0
FIND:
SOLUTION
For this arrangement,
δ E o GF
=
[ε 1 − ε 2 + ε
Ei
4
ε1 = εa + εb + εT
ε3 = -εa + εb + εT
4
−ε 3]
ε2 = -εa + εb + εT
ε4 = εa + εb + εT
Then,
δ Eo
Ei
= GFε
where the bridge constant κ = 4.
The relation between applied load, F, and strain is
F=
2E m Iε
Lt
or
F=
2E m Iε
Lt
rearranging,
δEo = 1.0V
=
2δE
o
AκGFE i
=
2 E m bt 2δE
o
3LAκGFE i
PROBLEM 11.25
KNOWN: It is desired to design a strain-gauge based scale, using a cantilever beam.
The beam is 21 cm long, 0.4 cm thick, and 2 cm wide. The loads are up to 200 g, applied
20 cm from the fixed end of the beam. The required uncertainty level is 4%.
FIND: Design a measurement system, including strain gauge placement, bridge
characteristics, and signal conditioning.
SOLUTION: Because this design problem has a wide variety of solutions, a general
approach and some representative equations and results will be provided.
The beam is made of 2024-T4 aluminum, having a modulus of 71 Gpa. For a cantilever
beam, the deflection at the free end is
b
=
f
Wl 3
bh3
=
where I
3Em I
12
h
Here f is the deflection, W is the load, l is the distance from the fixed end to the load, I is
the moment of inertia, and Em is the modulus. A representative design may be examined
by assuming a bridge having a single active gauge sensing the maximum axial strain. This
would correspond to a location on the surface of the beam at a location where the load is
applied. In this case the deflection f is
2
3
1960 )( 0.2 )
(
Wl 3
bh3 ( 0.004 )( 0.02 )
f
I
=
=2.7 cm
=
=2.67 ×10-9
=
9
−9
3Em I 3 ( 71×10 )( 2.67 ×10 )
12
12
Wl ( h 2 ) (1960 )( 0.2 )( 0.01)
=
= 1.47 ×109 Pa
−9
I
2.67 ×10
Then with Em = 71 GPa
σ x 1.47 ×109
δR
=
= 0.0207
= ε GF= 0.0207 × 2= 0.0414
ε=
9
71×10
Em
R
σx
=
δ Eo δ R R 0.0414
=
=
= 0.01
4
4
Ei
From these relationships, the output for a bridge excitation of 5 V is about 50 mV, which
would create an uncertainty from the A/D resolution of about 2.4%. The bridge can be
designed in a reasonable manner to meet the uncertainty constraint.
PROBLEM 12.1
KNOWN: A linear potentiometer having
D = 0.1 mm
ρe = 1.7 × 10-8 Ω-m
R = 1 kΩ
FIND:
a) for a core diameter of 1.5 cm, determine the range
b) plot loading error as a function of displacement
SOLUTION:
a) In order to determine the number of turns of wire, N
R=
ρe L
Ac
⇒L=
RAc
ρe
with
Ac =
π
c0.1 × 10 mh= 7. 854 × 10
4
−3
2
−9
m2
yields
b1000 Ω gc7. 854 × 10
L=
−9
m2
h= 462 m
1. 7 × 10−8 Ω - m
One turn takes π(1.5) cm of wire, thus
462 m
N =
= 9804
π 1.5 × 10-2 m
c
h
Then since each turn occupies approximately D = 0.1 mm the range is
(9804)(0.1 mm) = 0.98 m
b) The loading error for a voltage dividing circuit is found from (6.37) as
′
  Eo 
 −  
  Ei 
R − RT + (RT − R1 )(R1 / Rm + 1)
= Ei 1
 R2

RT +  T − RT (R1 / Rm + 1)
 R1


E
eL = Ei  o
 Ei

In terms of output voltage the length may be written, assuming an infinite meter
resistance, from (6.8)
L
R
E o = x Ei = x Ei
LT
RT
yielding
el
× 100 = loading error as a percentage of full - scale deflection
Ei
A plot is shown below.
14
1000 Ohms
10000 Ohms
100000 Ohms
12
% Loading Error
10
8
6
4
2
0
0
0.1
0.2
0.3
0.4
0.5
Length (m)
0.6
0.7
0.8
0.9
1
PROBLEM 12.2
KNOWN: Numerous applications for linear displacement sensors.
FIND: Develop specifications for a selected application for a linear displacement sensor.
SOLUTION:
A list of potential applications is provided in the problem statement. These applications
include several measurements that would easily be accomplished by a potentiometric or
LVDT position sensor having an accuracy of ±1 mm, including seat position and throttle
position sensors for automotive applications.
There exist applications where errors must be on the order of microns, and a variety of
sensors exist that can meet such a stringent repeatability requirement, although absolute
accuracy can be a significant challenge. Applications include machine tools and rapid
prototyping systems.
PROBLEM 12.3
KNOWN: Potentiometers are primarily either wire-wound or conductive plastic in their
construction.
FIND: Compare and contrast the design of wire-wound and conductive plastic
potentiometers.
SOLUTION:
Conductive plastic potentiometers provide continuous analog output, where wire-wound
potentiometers have resolution limited by the size of the wire that forms the winding.
However, production of conductive plastic having a constant resistivity, which produces a
linear relationship between displacement and resistance, is not perfect. Both designs are
widely applied in a variety of devices.
PROBLEM 12.4
KNOWN: LVDT and potentiometric displacement transducers.
FIND: Compare and contrast the use of LVDT and potentiometric displacement
transducers.
SOLUTION:
The LVDT has the following positive characteristics:
•
•
•
•
Because there is no contact between the movable core and the coil structure,
friction is extremely low, and provides for essentially infinite life
Truly analog behavior, so that resolution is limited only by the output measuring
system employed
Very good repeatability
Somewhat greater cost
The potentiometric sensor, has the following positive characteristics;
•
•
Lower cost
Good repeatability
PROBLEM 12.5
KNOWN:
y(t) = 0.2 cos 10t + 0.3 cos 20t
where y = displacement [in.]
and t = time [sec.]
ζ = 0.7
k = 1.2 lb/ft
FIND:
a) a combination of m and c which would yield less than 10% amplitude error in
measuring the input signal.
b) determine the phase response of the system
SOLUTION: We know
=
ωn
=
k m cc 2 km
The amplitude error is evaluated by examining (for cos = 1)
( yr ) steady
A
=
(ω ω n )
{
2
2
1 − (ω ω n ) 2  +  2ζ (ω ω n )  2


 
}
1
2
The limiting case is for the lower input frequency, and ( yr ) steady A = 0.9
0.9 =
{
10 
 ω 
n

2
2
1 − (10 ω n ) 2  +  2 ( 0.7 )(10 ω n )  2


 
Solving for ω n yields
ω n = 7.1 rad/s
and since
}
1
2
ωn =
kg c
m
and c is found from
c = 2ζ km
gc
yielding
1.2 lb/ft ( 0.766 lb m )
lb sec
0.236
=
c 2=
( 0.7 )
ft lb m
ft
32.174
2
lb sec
b) Phase response is shown below.
0
0
5
10
15
20
25
-20
Phase Angle (deg)
-40
-60
-80
-100
-120
-140
-160
-180
Omega (rad/s)
30
35
40
45
PROBLEM 12.6
KNOWN: A seismic instrument has
=
ω n 20
=
Hz 40π rad/s
ζ = 0.65
FIND: The maximum input frequency, ω for a vibration measurement such that the
amplitude error is < 5%.
SOLUTION:
From (12.14)
0.95 or 1.05 =
1
2 2
 
 


 1 −  ω ω   +  2ζ
n 


 
Solving for the input frequency yields 90.5 rad/s.

ω  

 
 ωn   

2
1
2
PROBLEM 12.7
KNOWN: A seismic instrument may be designed to measure either acceleration or
displacement (vibration).
FIND: Clearly state the requirements for the values of the spring constant, the damping
coefficient, and the mass in the seismic instrument to achieve the desired output.
SOLUTION:
For vibration measurements, the output of the seismic instrument must accurately provide
the amplitude of the displacements associated with the vibrations; thus, the desired
behavior of the seismic instrument would be to have an output that gave a direct indication
of yh. For this to occur, the seismic mass should remain essentially stationary in an
absolute frame of reference, and the housing and output transducer should move with the
vibrating object. To determine the conditions under which this behavior would occur, the
amplitude of yr at steady state can be examined.. Figure 12.9 shows ( yr )max / A as a
function of the ratio of the input frequency to the natural frequency. Thus, a seismic
instrument that is used to measure vibration displacements should have a natural frequency
smaller than the expected input frequency. Damping ratios near 0.7 are common for such
an instrument. The seismic instrument designed for this application is called a vibrometer.
On the other hand, the measurement of acceleration requires that the seismic mass relative
position is a direct measure of acceleration. This requires a magnitude ratio of unity, and
an appropriate natural frequency.
PROBLEM 12.9
KNOWN: A seismic accelerometer has:
m = 0.2 g
k = 20 000 N/m
Very low damping
FIND: Instrument bandwidth
SOLUTION:
The natural frequency is found as
=
ωn
k
=
m
kg m
2
sec=
m 10, 000 rad/sec
-3
0.2 × 10 kg
20, 000
and with a very low damping. The bandwidth may be found from Figure 3.16 with the
allowed frequency range from 0 to 0.4ω n or 0 to 4000 rad/s.
PROBLEM 12.10
KNOWN: Integration reduces the effects of noise in a signal.
A moving average integrates over a fixed time interval, based on a concept called
windowing.
Noise in the present case has significant amplitude, but a significantly higher frequency
than the measured velocity.
FIND: The effect of employing a moving average on a signal
SOLUTION: This signal has high frequency noise present.
3
2
Y(t)
1
0
-1
-2
-3
0
5
10
15
20
25
30
Time (sec)
A 5 point moving average has been performed on the signal, and much of the high
frequency noise removed.
1.5
1
Y(t)
0.5
0
-0.5
-1
-1.5
0
5
10
15
Time (sec)
20
25
30
PROBLEM 12.11
KNOWN: Moving coil transducer with
=
Dc 0.8
=
cm l 2 cm dy
from 1 to 10 cm/s
dt
A/D converter with 8-bit resolution and 0.1% FS accuracy
FIND: The required magnetic field strength as a function of N for 1% accuracy in the
velocity measurement.
ASSUMPTION: Assume that the uncertainty in the magnetic field strength and the
number of turns are negligible.
SOLUTION:
From equation 12.19, the emf from the moving coil can be expressed
dy
emf = π BDc lN
dt
dy
the uncertainty in the velocity can then be expressed, with V =
dt
uV uemf
=
V emf
The uncertainty in the emf has two contributions. From the resolution of the A/D, the
uncertainty is ± 7.8 mV plus 1 mV (RSS addition) from the accuracy, yields uemf =
7.9 mV. Combining the relations for emf and the uncertainty in V, with an uncertainty in
V of 1%, yields the plot below
PROBLEM 12.13
KNOWN: The development of MEMS and MEMS related applications has increased
the requirements for measuring very small forces.
FIND: Current state-of-the art in force measurement for very small forces.
SOLUTION:
Numerous journal articles and websites can serve as resources to assess the various limits
of force measurement. For biomedical applications, micro-load cells have been developed
with a range from 0 to 2 N. MEMS applications are on the order of 0.2 mN. And the
Atomic Force Microscope uses forces at the atomic level to provide surface topology.
PROBLEM 12.14
KNOWN: A proving ring is to be designed to serve as a calibration standard for forces
over the range from 250 to 1000 N.
FIND: A suitable design to provide reasonable uncertainty.
SOLUTION: It is necessary to establish the minimum deflection which can be
measured with reasonable accuracy. As an example consider a displacement transducer
having a range of 0 to 1 mm for an output of 0 to 5 Volts. If sampled using an 8-bit A/D,
the resolution would be 19.5 mV, corresponding to 0.004 mm. If we assumed that the
proving ring would deflect 1 mm at 1000 N, we can size the ring.
Assuming the cross section of the ring is rectangular, the moment of intertia is bh3 12 ,and
the deflection is given by
3
π 4  F D
δ=
y  −  n
 2 π  16 EI
Let’s assume that the ring is steel with a modulus of E = 20 × 1010 Pa. By varying the
cross section and the diameter, a suitable deflection can be established. As an example of
the process, assume a square cross section, and a diameter of 8 cm. A deflection of 1 mm
at 1000 N would be achieved with a dimension of 4.9 mm for the square cross section.
Then at a load of 250 N the deflection would be 0.25 mm, yielding an output voltage of
1.2 V, which should yield a reasonable uncertainty.
PROBLEM 12.15
KNOWN: Power transmission through a drive shaft results in 1800 rpm with a power
transmission of 40 hp
FIND: Torque transmitted by the driveshaft.
SOLUTION:
With
and P = 40
hp , ω
=
P = ωT
1800 × 2π
= 188.5 rad/s , we find that the torque is
60
T=
( 40 hp )( 550 ft-lb sec-hp ) 117 ft-lb
188.5 rad/sec
PROBLEM 12.16
KNOWN: A dynamometer can be an integral part of emissions testing for automotive
applications.
FIND: Define and discuss the importance of a dynamometer in automotive emissions
testing.
SOLUTION:
Automotive emissions vary as a function of load and speed. Engine temperature and the
temperature of the catalytic converter are also important parameters. Various levels of
sophistication are possible for the testing of IC engine emissions. An engine analyzer
combined with a dynamometer provides an extensive range of analysis capabilities. A
dynamometer will allow monitoring of engine rpm and power output, and may allow a
variety of measurements including throttle position, individual cylinder vacuum level,
spark quality, and fuel flow rate. Clearly, for emissions testing the engine exhaust is
routed through emissions testing equipment.
PROBLEM 12.17
KNOWN: Linear actuators form an important component for a variety of systems.
FIND: Research applications for linear actuators.
SOLUTION:
The following list represent applications for linear actuators that provide a wealth of
information online or in refereed journal publications:
•
•
•
•
•
Autonomous vehicles
Manufacturing
Pick-and place operations
Injection molding
Positioning in research applications requiring precision positioning
PROBLEM 12.18
KNOWN: Pneumatic cylinders provide linear actuation, primarily to yield only two
positions.
FIND: Research the range of displacement and force that can be provided by pneumatic
cylinders.
SOLUTION:
Representative ranges for pneumatic cylinders would be:
•
•
•
•
Bore sizes from 0.5 to 15 inches
Stroke length up to 24 inches
Operating pressure up to 500 psig
Single or double acting cylinders
PROBLEM 12.19
KNOWN: Flow through a control valve, at a flow rate of 32 SCFM, and an associated
pressure drop of 10 psi. The line pressure (the pressure at the inlet side) is 100 psig, and
the temperature and relative humidity are 68°F, 36%, respectively.
FIND: Research the range of displacement and force that can be provided by pneumatic
cylinders.
ASSUMPTIONS: The effect of relative humidity on the result would be very small,
and will be neglected in this analysis.
SOLUTION:
Continuity is applied to find the flow rate at the inlet of the valve in ACFM.
ρ1 ( AU )1 = ρ 2 ( AU )2
where 1 represents the standard conditions (T = 77°F, 1 atm) and 2 represents actual
conditions. Employing the ideal gas model, we compute the two densities as
=
ρ1
P
=
RT
(14.7 )(144=
)
( 53.34 )( 537 )
0.074 lb m ft 3
and
=
ρ2
P
=
RT
(100 + 14.7 )(144
)
=
( 53.34 )( 528)
0.586 lb m ft 3
Then with
ACFM =
ρ1
SCFM
ρ2
we find that the flow rate in ACFM is 4 CFM. The flow coefficient for the valve is
ACFM
= Cv ∆P
=
Cv
4
= 0.106
10 ×144
PROBLEM 12.20
KNOWN: A low profile control valve is employed for a filling and exhaust process.
The volume to be filled is 500 mL. During filling the valve has
F = 0.82 ms/cc, with a lag time of 8 ms . During exhaust, the valve has
F = 0.7 ms/cc, with a lag time of 8 ms
FIND: The time required for the filling and exhaust processes.
SOLUTION:
The time is given by
t90= m + FV
So for fillinf
8 ms + ( 0.82 ms/cc )( 500 cc ) =
418 ms
t fill =
and for exhaust
texhaust =
8 ms + ( 0.7 ms/cc )( 500 cc ) =
358 ms
PROBLEM 12.21
KNOWN: Research the design of a totalizing flow meter.
SOLUTION: Totalizing flow meters currently allow for remote reading of meters, and
subsequent billing in many applications. For most residential applications, positive
displacement meters allow for very accurate metering even at very low flow rates; designs
include disc meters with either oscillating disk or piston. At commercial installations,
turbine or propeller type flow meters are often employed. See also Plumbing Systems and
Design, July/August 2003 pp. 70.
PROBLEM 12.23
KNOWN: Research the design of a residential thermostat.
SOLUTION:
The two key elements of a mechanical thermostat are a mercury switch and a bimetallic
thermometer, as shown below. The mercury switch is oriented so that electrical contact is
created when the mercury flows from one end of the glass bulb to the other. The
deadband is created by the angle of the bulb relative to gravity required to overcome
friction and allow the mercury to flow.
PROBLEM 12.24
KNOWN: Show that a proportional controller has a steady state error. Develop an
expression for the steady state error when a proportional controller acts on a first-order
plant.
SOLUTION:
Let the desired value of the controlled variable be represented as d ( t ) , and the error
signal be e ( t ) , with the Laplace transforms of these two quantities represented as
D ( s ) and E ( s ) , respectively. Then with the plant transfer function G ( s ) , the feedback
transfer function H ( s ) , we can state
E (s) =
D (s)
1+ H (s)G (s)C (s)
where C ( s ) represents the control function. For proportional control, C ( s ) =Kp. And
from the final value theorem,
steady-state error =
1
.
1+G ( s ) K p